document resume cs 208 165 - eric · 2014-03-30 · document resume. ed 240 60?, cs 208 165. author...
TRANSCRIPT
DOCUMENT RESUME
ED 240 60?, CS 208 165
AUTHOR King, Martha L.; Rentel, Victor M.TITLE Transition to Writing. Final Report. Volume II.INSTITUTION Ohio State Univ., Columbus. Research Foundation.SPONS AGENCY National Inst. of Education (ED), Washington, DC.PUB DATE Sep 82GRANT NIE-G-79-0039; NIE-G-79-0137NOTE 274p.; For Volume I of this series, see ED 213
050.PUB TYPE Reports - Research/Technical (143)
EDRS PRICE MF01/PC11 Plus Postage.DESCRIPTORS Child Development; Child Language; *Cohesion (Written
Composition); *Developmental Stages; Grade 1; Grade2; Kindergarten; *Language Acquisition; LongitudinalStudies; Primary Education; *Story Telling; Syntax;*Writing Readiness; *Writing Research
IDENTIFIERS *Story Structure
ABSTRACTPresenting a longitudinal study of factors
influencing the text-forming strategies children employ in earlystages of writing development, thii report focuses on the differencesbetween children's oral and written texts and the development ofwriting ability. The first two chapters present cohesion results,with the firstchapter providing dictation and writing data for thegrades 1 and 2 population and the second chapter presentingreplication results comparing kindergarten-grade one and olderpopulation (grades 1-2) groups' cohesion in both dictation andwriting. The third chapter discusses story structure results in termsof their theoretical significance and implications for writingdevelopment, while a fourth chapter integrates cohesion and storystructure conclusions into an analysis of the transition to writingemphasizing patterns of development. Finally, the fifth chapterpresents a case study of the development of a kindergarten child'swriting abilities. Appendixes include (1) a description of studyprocedures, (2) cohesion MANOVAs by mode and grade for urban andsuburban schools, (3) cohesion MANOVAs for school, mode, andobservation, (4) story structure MANOVAs in retelling dictation andwriting at urban and suburban schoolse and (5) a discussion of textlength and syntactic complexity. (HOD)
***********************************************************************Reproductions supplied by EDRS are the best that can be made
from the original document.***********************************************************************
;
a
C
er
11
2-
U.S. DEPARTMENT Off EDUCATIONNATIONAL INSTITUTE OF EDUCATION
EDUCATIONAL RESOURCES INFORMATION' CENTER IERICI
Atha document has been reproduced asreceived from the person or Orgon0600ononoon9 II
Moot changes have been made 10 noneoreproduction 0000
Points of *no co moons stated in thedoCloment do 001 nocomolly roposemotheal NIEposition d policy
thioohlovat*univisity
C
Ri Project 761861/712383and 761513/711748
Final ReportVolume II
roseatich foundation1314 Mimi's' rood
coluirobus, ohlo.43212
TRANSITION TO WRITING
MARTHA L. KING AND VICTOR Mg RENTEL.
The College of Education1945 N. Nigh Street
Collagist, Ohio
DEPARTMENT OF HEALTH AND HUMAN SERVICESNational Institute of Education
Washington, D.C.. 20205 m.
Crint No. NIE-G-79-0137and
GrantNo. NTH-G-79-0039
Septaiber,. 1982
Acknowledgements
We wish to express our deep appreciation to Christine Pappas andBarbara Pettegrew who coded the cohesion data and, more importantly, whosecritical comments and suggestions contributed importantly to the substanceof this research. We also owe a debt to Jolaine Scholl who programmed ouranalyses and gave us invaluable statistical assistance at key points inour analyses. We Wish also to thank Ellen Martin Huff, Barbara Kiefer,John Quinn and Lyn Zalusky who assisted with story coding. Our specialgratitude goes to the teachers and pupils at Barrington Informal School inUpper Arlington, Joetta Beaver, Nancy Blume, Marlene Harbert, SherlynFernandez, Sandra Saunders, and Marilyn Reed, Principal; and those atDouglas Alternative School in Columbus, Sandy Barthelmas, Mary Bornstein,Connie Cline, Denise Harrison-, Bruce Stassfurth, Mary Sykora, and Kay Noble,Principal. We wish also to acknowledge the support of the Columbus PublicSchools and the Upper Arlington Public Schools.
Very special recognition is due Rugaiya Hasan. Professor Hasan'sassistance as a consultant in the early stages of our work was instrumentalin resolving a host of issues and problems. Also John Kennedy deservesspecial thanks for backstopping our design and analysis decisions. And, toC. Ray Williams, Chairman of Early and Middle Childhood Education, and GilJarvis, Chairman of Humanities Education, who somehow managed to findresources when we needed them, our everlasting gratitude. Finally, vewant to thank Gay Pinell-WaYson for her assistance in collecting data atthe outset of the study.
MIX VMR
Preface
Volume 1 of this series of longitudinal studies dealt with the
contribution of a variety of factors which influence the text-forming
strategies children employ in early stages of writing development.
Constitutional, socioeconomic, and maturational factors were described
which affected distributions of cohesive ties and the range, sustaining'
power, and complexity of their written stories. Thlevolume focuses
on differences between children's oral and written texts and on deyelop-
ment of writing ability --in other words, on their transition to writing.
Although the same constitutional, socioeconomic, and maturational variables
are considered briefly, further analyses have not uncovered substantial
new findings about these population characteristics. Readers of the present
volume who wish to, examine these differences in detail will find related
analyses in Appendices B, C and D. Procedures may be found in Appendix A.
The present volume'is organized as follows. Cohesion results are
presented in the first two chapters. Chapter 1 extends Volume 1 by adding
dictation and writing data for the Grade 112 population, Chapter 2 presents
replication results for the kindergarten/grade-one groups compared with the
older population (Grade 1-2) for cohesion in both dictation and writing.
Story structure results, as in Volume 1, are given in a separate
chapter (3). For both cohesion and story structure major differences in
speech and writing are discussed in terms of their theoretical significance
and implications for writing development. In a fourth chapter, cohesion
and story structure conclusions are integrated into an analysis of the
transition to writing emphasizing patterns of development. Finally, a
case study of a kindergarten child is presented in, Chapter 4.
MLK VMR
1
1
Table of Contents -
Chapter Page
Acknowledgements ii
Preface
1 The Transition to LiteracyMAKING MEANINGS EXPLICIT
Contextual Factors 1
Making Meanings Explicit in Writing 2
Cohesion in Children's Oral and Written Texts:Summary Grade 1-2 Results 4
Transition to Writing: From Dictation to WrittenText Results for Grade 1-2 6
Multivariate and Univariate Analyses 7
Overviev of Cohesion Results: Comparison of Dictatedand Written Modes, Grade 1-2 8
Results of MANOVA for Dictation and Writing '9
Urban School Cohesion Results: Dictation and Writing 10
Observation and mode factor follow-ups 15
Dialect by mode interaction and dialect factor follow-up -16
Sex factor follow-up 20
Urban School Cohesion Results: Dictation Only 21
Dialect by sex interaction follow-up 22
Observation factor follow-up 24
Suburban School Cohesion Results: Dictation and Writing 25
Mode by observation interaction and observation follow-ups 26
Sex by mode interaction follow-up in dictation andwriting 29
Suburban School Cohesion Resuits: Dictation Only 32
2 Kindergarten/Grade One Replication of the Transitionfrom Speech to Writing 34
Cohesion Results for Dictation and Writing:Replications (K-1/1-2) 35
Urban School Cohesion Results for Dictation:Replication (K-1/1-2) 35
Observation factor follow -up 38
Urban School Cohesion Results for Writing:Replicati.pn (K-1/1-2) 40
Grade by dialect interaction in writing at urban school 40
Grade by sex interaction in writing at urban school(K-1/1-2)--follow-up 43
Grade factor follow-up 47
Dialect factor follow-up 47.
Observatiqn factor follow-up 48
Suburban School Cohesion Results for Dictation:Replication (K-1/1-2) 49
Suburban School Cohesion Results for Writing:Replication (K-1/1-2) 50
Grade by observation follow-up 53
iv
Chapter
2 Continued
Table of Contents (Continued)
Page,
Sex by observation follow-up 55
Sex factor follow-up 57Observation factor follow-up 57Discussion of Cohesion Results 58The Transition to Writing: Mode and Grade Comparisons 58Restricted Exophoric Reference 59.
Reference 60Ellipsis 61
Conjunction 62
Lexical cohesion 63
3 The Transition to Writing: Story Structure forGrade 1-2 w 66
Analysis of Story Structure Data 72
Structure: Mode Comparisons 76StoryFollow-Up ANOVAs For the Discourse Contexts Factor 79
Follow-Ups ANOVAs for the Observation Factor 81
1Observation Interaction 82 11
Follow-Up ANOVAs for the Discourse Context by
The Retelling Context 83
The Writing Context 84
Follow-Up ANOVA on Functions 85
Follow-Up ANOVA on Function Types 87
Follow-Up ANOVA on Moves 87
The Dictation Context 87
Are Stories Relevant Production Models forBeginning Writers? 88
4 The Transition to Writing 94
Developmental Aspects of the Transition to Writing 97Factors Contributing to Developmental Variation in
the Transition to Writing 102
School variation 103
Sex variation 105.
Dialect/socio-economic status variation 105
Longitudinal Studies of the Transition to Writing 107
5 Transition to Writing: A Case Study
The Development of Text CohesionStory RetellingStory DictationAssigned WritingSummary.
ThelSavelopment of Story FunctionsStory RetellingStory Dictation
112
113113116
117
120
121
121
122
Table of Contents (Continued)
Chapter
5 Continued
KIM
Assigned Writing , 122
Summary 122
Unassigned Writing 123
Attitudes Abbut Writing 123
Context for Writing 124
Conclusion 125
BIBLIOGRAPHY.
Appendices:
127
A Procedures of the Study 131
B Cohesion MANOVAs by Mode and Gradefor Urban and Suburban Schools 155
C Cohesion MANOVAs for School, Modeand Observation (K-1) 168
D Story Structure MANOVAs in RetellingDictation and Writing at Urban (K-1)
and Suburban Schools (K-1) 190
E Text Length.and Syntactic Complexity 211
Additional Appendices in Volume 1.
Modified-Retelling
7
List of Tables
Table Page
1 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) in Dictation and Writingat Urban School (1-2) by Mode and Observation 11
2 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) in Dictation at the 'UrbanSchool (1-2) by' Dialect and Sex 12
3 Cohesion MANOVA By Dialect, Sex, Mode,* andObservation for Urban School (1-2) 13
4 Discriminant Analysis and Univariate ANOVAs asFollow -up to Significant Mode by ObservationInteraction in Cohesion MANOVA for Urban School (1 -2) 14-
5 -.Discriminant Analysis and Univariate ANOVAion Use ofCohesion Categories in Dictation and Writing forObservation at Urban School (1-2) 16
6 Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation,and Writing ior Mbde.at Urban School (1-2 17
7 ' Discriminant Analysis and Univariate ANOVAs as Follow-up
. to tignificant Dialect by Mdde Interaction in CohesionMANOVA for the Urban School' (j -2)
8 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) at the Urban School (1-2) by.Dialect and Mode
9 Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories' in Dictation and Writing forDialect at Urban School (1-2)
17
19
20.
10 Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation and Writing for Sexat Urban School (1-2) 21
11 Means and StandardDeviations for Cohesion MANOVA(Transformed Variables) at the Urban School (1-2) by Sex 21
12 Cohesion MANOVA in Dictation, by Dialect, Sex, and22
13 Discriminant Analysis and Univariate ANOVAs as Follow -upto Significant Dialect by Sex Interaction in CohesionMANOVA in Dictation for Urban School (1-2)
14 Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation for Observation at theUrban School (1-2) 24
Observation, for Urban School (1-2)
23
8
List of Table- (cont.)
Table Page
15 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) in Dictation at Urban School(1-2) by Observation 25
.
16 Cohesion MANOVA by Sex, Mode, and Observation inDictation and Writing for Suburban School (1-2)
17 Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation and Writing for Modeby Observation Interaction at Suburban School (1-2)
18 Means and Standard Deviations for CohesiOn MANOVA(Transformed Va:iables) at the Suburban School (1-2)by Mode and Observation
26
27
28
19. Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation and Writing for.Observation at Suburban School (1-2) 29
20 Discriminant Analys/s and Univariate ANOVAs on Use ofCohesion Categories in Dictation and Oriting for Sex byMode Interaction at SuSurbadSchool (1-2)
21 Means and Standard Deviations for Cohesion MANOVA by(Transformed Variables) by Sex and Mode inDictation and Writing at Suburban School (1-2) 31
22 Cohesion MANOVA in Dictation by Sex and'Observationfor Suburban School (1-2) 32
;3 Discriminant Analysis and Univariate ANOVAs in Dictationfor Observation Factor at Suburban School (1-2) 33
24 Cohesion MANOVA by Grade, Dialect, Sex, and Observationin Dictation for Urban School (K-1/1-2) 36
30
25 Means and Standard Deviations of Cohesion MANOVA(Transormed Variables) in Lictation for Urban School(IC 1/1-2)--by Grade and Observation 37
26 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inDictation for Observation Factor at Urban School (K-1/1-2) 38
27 Discriminant Function Coefficients and Univariate ANOVAson Use of Cohesion Categories in Dictation for Grade byObservation In naction at Urban School (K-1/1-2) 40
28 Cohesion MANOVA by Grade, Dialect,. Sex, and Observationin Writing Urban School (K-1/1-2)
0 9
viii
41
List ofTables (cont.) '
Table Page
29 Discriminant Function Coefficients and UnivariateANOVAs on Use of Cohesion Categories in.Writing forGrade by Dialect Interaction at Urban School (K-1/1-2) 42
30 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) on Writingl.or Urban School(K-1/1-2)--by Grada and Dialect
31 Discriminant Function Coefficients and UnivariateANOVAs on Use of Cohesion Categories in Writing forGrade by Sex Interaction at Urban Schoolv(K-1/1-2)
44
45
32 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) on Writing for Urban School(K-1/1-2)--by Grade and Sex 46
33 - 'Discriminant Function Coefficients and UnivariateANOVAs on Use of Cohesion Categories in Writing forGrade Factor at Urban School (K-1/1-2.) 47
34 Discriminant Function Coefficients and UnivariateANOVAs on Use of Cohesion Categories in Writing forDialect Factor at Urban Schbol (K-1/1-2) - 48
35 Discriminant Function Coefficients and Univariate ANOVAson Use of Cohesion Categories in Writing for ObservationFactor at Urban School (K1/1-2) 49
36 Means and Standard Deviations of Cohesion MANOVA(Transformed Variables) on Writing for Urban School
At(K-1/1-2)--by Observation 49 .
11/
37 Cohesion MANOVA by Grade, Sex, and Observation inDictation for the Suburban School (K-1/1-2)
38 Cohesion MANOVA by Grade, Sex, and Observation inWriting for the Suburban School (K4-1/1-2)
39 Means and Standard Deviations of Cohesion MANOVA(Transformed Variables) in Dictation and Writing forSuburban School (K-0-2) by Grade, Mode, and 'Observation
40 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inWriting for Grade by Observation at Suburban School(K-1/1 -2)
41 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inWriting for Sex by Obeervation interaction at
Suburban School (IC-1/1-2)
50
51
52
53
SS
List of Tables (cont.)
Table Page
42 Means and Standard Deviations of Cohesion MANOVA(Transformed Variables) for Suburban School inDictation and Writing (K- 1/1 -2) by Sex, Mode, and
Observation 56
43 Standard Discriminant Function Coefficients andUnivariate'ANOVAs on Use of Cohesion Categories inWriting for Sex at Suburban School (1C1/1-2) 57
44 Standard Discriminant Function Coefficients and .
Univariate ANOVAs on Use of Cohesion Categories inWriting for Observation at Suburban School (1C1/1-2)
45 Two-Factor Repeated Measure Story Structure MANOVAby Discourse Contexts and Observations
46 Means and Standard Deviations of Functions, FunctionTypes and Moves by Discourse Contexts and Observations
47 Follow-up Univariate ANOVAs on Functions, FunctionTypes, and Moves by Discourse Contexts
48 Means and Standard Deviations of Functions, FunctionTypes, and Moves by Observations
49 Follow-up Univariate ANOVA on Function Types byObserliations
50' Story Structure MANOVA in Retelling Context byObservations
51 Means and Standard Deviations for Function:Types in
11
Retelling Context by Observations
52 Follow-up Univariate ANOVA in Retelling Context onFunction Types by Observations 84
58
78
79
80
81
82
83
53 Story Structure MANOVA in Writing Context byObservation 85
54 Means and Standard Deviations of Functions, FunctionTypes, and Moves in Writing Context by Observations' 85
55 Follow-up Univariate ANOVAs on Functions, FunctionTypes and Moves in Writing Context by Observation 86
56 Story Structure MANOVA in Dictation Context byObservations 87
11
List of Tables
APPENDIX A
Table Page
1 ANOVA of Socio-Economic Class by School and Sex 137
2 Means and Standard Deviations of Socio-EconomicClass by School and Sex 137
3 Number of Functions, Types, and Moves in ThreeStories 152
APPENDIX B
1 Cohesion MANOVA by Grade, Dialect, Sex, Mode, andObservation in Dictation and Writing at Urban School(K-1/1-2)
Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) for Urban School (K-1/1-2)--byGrade, Dialect, nnd MOdio.
3 Means and Standard Deviations of Cohesion MANOVA .
(Transformed Variables) in Dictation and Writing forUrban School (*-1/1-2)--by Grade and Observation
4 Mgans and Standard, Deviations of Cohesion MANOVA(Transformed Variables) for Urban School (1C1/1-2)--byGrade and Mode
5 Means and Standard Deviations of Cohesion MANOVA(Transformed Variables) for Urban School (K-1/1-2) - -by
Dialect and,Mode.
6 Means and Standard Deviations of Cohesion MANOVA(Transformed Variables) for Urban School (K-1/1 -2) --byGrade and. Sex
7 Means and Standard Deviations of Cohesion MANOVA(Transformed'Variables) for Urban School (K-1/1-2)--byMode and Observation
157 &158
159
.160
161
162
163
164
8 Cohesion MANOVA by Grade, Sex, Mode, and Observationin Dictation and Writing at Suburban School (K-J/J-2) 165
9 Means and Standard Deviations'of Cohesion MANOVA(Transformed Variables) in Dictation and Writing forSuburban School - -by Grade and Observation
10 Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) in Dictation and Writing forSuburban School - -by Grads, Sex, and Observation
166
167
1
1'
in
List of Tables (cont.)
APPENDIX C
Table
1 Cohesion MANOVA in Dictation and Writing by School,Sex, Mode, and Observation--(k-1)
Page
173
2 Cohesion MANOVA in Dictation by School, Sex, andObservation--(K-1) 174
3 Cohesion MANOVA in Writing by School, Sex, andObservation.- -(KA) 175
4 Standard Discriminant Function Coefficient. andUnivariate ANOVAs on Use of Cohesion Categorie3 inWriting foe School by Sex by Observation- -(K -1) 177
5 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inWriting for Observation--(K-1) 177
6 Means and Standard Deviations of Cohesion Categoriesin Writing by School, Sex, and Observation- -(K -1) 178
7 Cohesion MANOVA in Dictation and Writing by DialectSex, Mode, and Observation--at Urban School (K 1) 179
8 Cohesion MANOVA in, Dictation by Dialect, Sex, andObservation - -at the Urban School (L-l) 180
9 . Standard Discriminant Function Coefficients andUnivariate ANOVAs can Use of Cohesion Categories inDictation for Observation at Uxban School (K-1) 181
10 Means and Standard Deviations of Cohesion Categoriesby Mode and Observation at the Urban School (K-1) 182
11 Cohesion MANOVA by Dialect, Sex, and Observation inWriting for Urban School (KA) 183
12 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inWriting for Dialect at Urban School (K-1) 184
13 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories in(biting for Observation at Urban School (KA) 184
14 Means and Standard Deviations of Cohesion categories
in Writing by Dialect at Urban School (K-1,
15 CohesionMANOVA in Dictation and Writing by Sex, Mode,and Observatiow-Suburlan School (K-1)
185
- 186
List of Tables (cont.)
APPENDIX C (cont.)
Table
16 Cohesion MANOVA in D ation by Sex and Observationat the Suburban Sqh cl (K-1)
17 Cohesion MANOVA by Sex and Observation in Writingat Suburban School (K-1)
18 Standard Discriminant Function Coefficients andUnivariate ANOVAs on Use of Cohesion Categories inWriting for Observation at Suburban School (K-1)
Page
187
187
188
19 Means and Standard Deviations of Cohesion Categoriesby Mode and Observation at the. Suburban School (K-1) 69
11
APPENDIX D
111 Story Structure MANOVA in Dictation and Writing
by Dialect, Mode and Observation at Urban School(1C1) 195
2 Discriminant Function Coefficients and UniviriateANOVAs as Follow-up to Mode Effect for Urban School (KI) 196
3 Discriminant Function Coefficients and UnivariateANOVAs as Follow-up to Observation Effect for UrbanSchool (K -1) 196
4 Means and Standard Deviations of Story Structure Datafor Urban School by Mode and Observation
'5 Story Structure MANOVA in Dictation by Dialect, Sex,and Observation--for Urban School (K-1)
6 Discriminant Function Coefficients and UnivariateANOVAs as Follow-up to Observation Effect in Dictationfor Urban ScLrool (K -1) --Story Structure
7 Story Structure MANOVA in Writing by Dialect, Sex, andObservation- -for Urban School (K-1)
8 Story Structure MANOVA in Retelling by Dialect, Sex,and Observationfor Urban School (K -1)
9 Means and Standard Deviations for Function Types andFunctions at Urban School (KI) by Dialect andObservation
197
198
199
200
201
202
10 Discriminant Function Coefficients and UnivariateANOVAs as Follow-up to Dialect by Observation Interaction inRetelling for Ueoan School (K -1) --Story Structure 203
14
List of Tables (cont.)
APPENDIX D (cont.)
Table Page
11 Story Structure MANOVA by Sex, Mode, and Observationfor Suburban $,hool (K-1).--Dictation and Writing 205
12 -Story Structure MANOVA in Dictation by Sex andObservation --for Suburban School (K-1) 206
13 Story Skructure MANOVA in Writing by Sex andObservation -- Suburban School (K-1) 207
14 Story Structure MANOVA in Retelling by Sex andObservation--for Suburban School (K-1) 208
13 Discriminant Function Coefficients and UnivariateANOVAs as Follow-up to Observation Effect in Retellingfor Suburban School (K.1)--Story Structure 209
16 Means and Standard Deviations of Story Structure Datain Retelling for Suburban School by Observation (K-1) 210
APPENDIX'E
1 Means and Standard Deviations of Text Length in Dictationand Writing for both Schools (K1) - -by Mode andObservation 213
2 ANOVA of Text Length in Dictation and Writing bySchool, Sex, Mode, and Observation (1C1) 214
3 Means and Standard Deviations of Text Length inDictation and Writing at Urban School (K- 1)--by Sex,Mode and Observation 215'
4 ANOVA of Text Length in Dictation and Writing at theUrban School--by Dialect, Sex, Mode, and Observation 216
5 Means and Standard Deviations of Text Length in Dictationand Writing by Sex, Mode, and Observationat SuburbanSchool (K-1) 219
6 ANOVA of Text Length in Dictation and Writing at theSuburbax School by Sex, Mode, and Observation (K-1) 220
7 ANOVA of Text Length in Dictation by School, Sex,and Observation {K -1) 222
8 ANOVA of Text Length in Dictation by Dialect, Sex,and Observation at the Urban School (K-1)
xiv
15
223
List of Tables (cont.)
APPENDIX E
Table
9 ANOVA of Text Length in Dictation at Suburban School(10-1) by Sex and Observation
10 ANOVA of Syntactic Complexity in Dictation and Writingby School, Sex, Mode, and Observation (K-1)
Page
224
225
11 Means and Standard Deviations of Syntactic Complexityin Dictation and Writing by School, Sex andObservatioi. CC -1) 226
12 Means and Standard Deviations of Syntactic Complexity(K- 1)--by Mode and Observation 229
13 ANOVA of Syntactic Complexity in Dictation and Writingat the Urban School (K-1) by Dialect, Sex,Mode, and Observation 230
L4 Means and Standard Deviations of Syntactic Complexity inDictation and Writing at Urban School (kg) by Dialectand Sex 231.
15 Means and Standard Deviations of Syntactic Complexity inDictation and Writing at Urban School (K11) byObservation 232
16 ANOVA of Syntactic Complexity in Dictation and Writingat Suburban School by Sex, Mode, and Observation (-.1) 233
17 Means and Standard Deviations of Syntactic Complexity inDictation and Writing by Sex, Mode, and Observation atSuburban School (K-1) 234
Figure
1
List of Figures
Observation as a Function of Mode for RestrictedExophoric Reference, Conjunction, and LexicalCohesion at Urban School (1-2)
2 Dictation and Writing as a Function of Dialect at
Urban School (1-2) 18
3 Reference and Conjunction as a Function of Dialect inDictation at Urban School (1-2) 24
Page
14
4 Lexical Cohesion and Reference in Dictation and .4Writing as a Function of Observation (1-2) 27
5 Ellipsis as a Function of Sex in Dictation and Writing
at Suburban School (1-2) . 30
6 Observation as a Function of Grade in Dictation at Urban
School (K- 1 /l -2) 39
7 Dialect as a Function of Grade for Ellipsis and LexicalCohesion in Writing for Urban School Mg/1-2Y 42
8 Observation as a Function of Grade for RestrictedExophoric Reference, Reference, Ellipsis, Conjunction,and Lexical Cohesion in Writing at Suburban School(K-1/1-2)
9 Observation as a Function of Sex for Reference inWriting at Suburban School (K-I/1-2)
10 Retelling of Squawk to the Moon Little Goose
11 Retelling of The Magic Porridge Pot
12 Retelling of Salt
13 Dictation: Spring-Kindergarten; Spring -Grade One
14 Dictation: Autumn-Grade One
15 Writing: Spring-Kiidergarten
16 Writing: Autumn-Grade One
17 Writing: Spring-Grade One
17
54
55
113
114
115
116
117
118
119
120
List of Figures (cont.)
Figure
A 1 Notational System for Editing Oral LanguageTranscripts
A 2 Cohesion Multivariate Analyses of Variance
Page
144
146
C 1 Observation as a Function of Sex and School for Schoolby Sex by Observation Interaction in Retelling (K-1) '176
D 1 Observation as a Function of Dialect at Urban School(K-1) for Function Types and Functions 204
E 1 Observation as a Function of Sex and Mode on TextLength for Urban School (K -1) 217
E 2 Mode as a Function of Sex on Text Length for SubUrban
School (1C1) 221
E 3 Observation as a Function of Sex add School in Diitationand Writing (K-1) 227
E 4 Mode as a Function of Observation in Dictation and
Writing (K1) 229
E 5 Dialect as a Function of Sex in Dictation and Writing
at Urban School (K-1) 231
E 6 Mode as a Function of Observation at Urban School (K-1) 232
_
18
Chapter 1
The Transition to Literacy
MAKING MEANINGS EXPLICIT .
Contextual Factors
8y school age, children are skilled, if not expert speakers. They have
learned to use language as a functional extension of mind and self, largely
within familiar, intimate, face-to-face situations where language may even
be aucillary to on-going actions (Bruner, 1975; MacNamara, 1972). The pur-
pose of these actions, the nature of the situation and its participants, and
the social setting itself are subtle natural conditions which children have
learned to interpret and address when they speak (Halliday, 1978). They
have learned that the situation itself is part of the meaning they must
understand and convey.
The situation may serve as the impetus for musts of what is said. From
the nature of.the activity and those who participate in it, various Rinds of
discourse are required. Interpersonal relations must be maintained. Necessary
information must be provided to solve problems or accomplish tasks. Coopera-
tion may be needed to maintain or complete the activity and, of course,
differences may arise among participants at any juncture where motives or
preferences differ. All of these and many other factors support and sustain
spoken discourse.
In most contexts where speech is applicable, participants in the discourse
usually share a wide range of knowledge about the situation and about each
other. This shared knowledge allows speakers to refer to objects, actions,
states, and persons using a range of communicative capabilities. They may
Pr
J
point physically, or by the very nature of shared attention to the task at
hand, they may simply rely on the perctptual salience of the activity to
carry the burden of reference. Some social activities symbolize rich cultural
meanings shared implicitly by members of the culture. Other activities
symbolise deeply personal meanings shared by just two people. In either case
speech may be ancillary to action or even ancillary to nonverbal meanings
conveyed by the situation. In either case, context permits both verbal and
nonverbal transmission of a range of meanings from highly implicit to highly
explicit (Masan, In press). Children, as participaUts, have learned to use
the varied codes that mediate meaning in a context of situation, each code
overlapping to some extent, but each code uniquely suited to convey certain 11
kinds of meanings (Argyle and Kendon, 1967). They have learned to talk using
the sum of these available codes, Aach adapted to accompany and interact with
the others, speech included. Speech is part of this tapestry of meanings.
Making Meanings Explicit in Writing -
In writing, where the text itself is the relevant environment for :stab
lishing all meaning relations, both attention and intention must be realigned.
Unlike speech,here attention may be directed always to intention and con-
text, attention must be directed as well to means and the form of language
(Cazden, 1974). To free their language from situational constraints, children
must understand that the text itself is the relevant environment for estab-
lishing meaning relations and learn also the unique features of written
texts (Masan, 1973; Pettigrew, 1981). They must learn to make meanings clear
within a text and :o direct language to a distant, often abstract audience
(Donaldson, 1978; Francis, 1975). What children learning to write must grasp
is how to. take what is implicitly obvious in context-of-situation and make
it explicit in written text (Cook-Gumperz, 1977). They must learn to
-2- 20
1
1.
appreciate language as a structure separate from action, placing increased
reliance on syntactic and semantic "foregrounding" as dominant carriers of
meaning--in short, to substitute words and syntax for gesture, prosodic
information, attributes of the discourse setting and all the redundant,
sources of meaning inherent in conversation (Cook-Gumperz, 1977; Doughty,
Pearce & Thornton, 1972; Ure, 1971).
Children also must learn the functions of written discourse. A. most
important distinction between speech and writing can be made on the basis
of function (Halliday, 1973). Halliday argues that spoken language essen-
tially has an interpersonal function. Spoken language is used predominantlyft
to maintain social relations. Written language' mainly serves.= ideational
function. This function manifests language's capacity toexpress the content
of experiences and their logical relationships. Olson (1977) also distin-
guishes between utterance and text. He, like Halliday, makes this distinc-
tion on the basis of function. Olson argues that utterances serve largely
to maintain social relations while texts specify truth functions. Utterances
have many implicit dimensions but texts are explicit: they explain and
describe rather than regulate and maintain social relations. Texts are
statements coded for reflection rather than for action., Halliday (1973)
has defined this specialized character of texts as language's textual function.
In writing, children must attend not only to ideas, but how ideas are expressed
the textual function.
Texts, (stories, pain's, arguments, explanations) are units of meaning
encoded in sentences (Halliday & Risen, 1976). Texts have meaning within
themselves and in relation to the context of which they are a part. The
production of a text is shaped by a setting "relevant actions and events,
by relationships among participants in a discourse, and by the mode of
-31' 21
communication employed. Texts are internally consistent; they comprise a
kind of unity. This integration or unity of meaning is achieved by the way' -
certain linguistic elemepts are us.ed. Halliday and Hasan (1976) have shown
how this unity is achieved through cohesion: that is, by establishing
semantic relations through interpretive elements in a text. Cohesion is
the range of theseinterpretive relations or ties. Ties are achieved
through five distinct kinds of cohesive relations: reference, substitution,
ellipsis, conjunction, and lexical cohesion (Halliday & Masan, 1976; King
& Rentel, et'al., 1981). Another task children must accomplish in learning
to write is to realign their use of cohesive ties to create texts which
stand independent of contextual props, yet maintain a unity of meaning.
COHESION IN CHILDREN'S ORAL and WRITTEN TESTS:SUMMARY GRADE 1-2 RESULTS
This report is the second of two designed to explore the kinds of shifts
in language usage children must make as they move from oral to written dis-
course. In the first study which began with children .midway through first
grade, we found that over the first 16 months of the study, from February of
first grade through all of second grade, children increased their use of
lexical, cohesion dramatically irrespective of social or linguistic background.
On the average, children increased the proportion of lexical cohesion in
writing from roughly 18 percent to nearly 50 percent of cohesive ties employed.
In addition to lexical cohesion, children used conjunctions proportionately
more, and used a wider range of them, adding approximately five percent more
conjunctions to their texts every four months. Words were used to link
meanings both at the level of proposition, and at higher levels of schema.
For example, propositions were linked in beginnint., texts, like the one that .
follows, by additive conjunctions, repeating words and using synonyms. Note
also that in line (3), the ultimate referent for they lies outside the text,
ftbut in (2), they refers back to bomb 3.
(1) Once there were these bombers.
(2) And they tried to destroy this bridge.
(3) And they had airplanes and bombers.
Such use of reference, both exophoric and endophoric, reference outside and
within the text, marked children's early productions. About 34 percent of
the total ties in children's beginning texts referred to something or some-,
one outside the text, but this percentage declined swiftly to three percent
by the second month of grade two and remained at that level. throughout the
year. Thus, children, regardless of class, sex, dialect, or school increas-
ingly confined reference ties to their texts. .Both substitutiod and ellipsis
were emploted sparingly.
These overall developmental differences in use of cohesion are illustrated
by the text below. This excerpted text is from a sample of writing collected'
at the end of second grade. The excerpts are from a much longer text but are
presented with intervening units omitted to convey a sense of how various
cohesive devices are employed over large spans of text to link presupposed
meanings.
(1) Once there was a little hamster named Dancer.
(8) Dancer ran all around the house.
(9) Then someone opened the door.(10) She skitter-skattered out of the house.
(14) When she was outside, she made lots of friends.
(18) Toby, the tomcat, was her best friend.(19) She met Bola Bom, the bird, and Tommy, the tiger.
First, there are the lexical ties: around, out, outside (collocation); house,
door (meronymy or part-whole relationships); ran, slexter-skattered (synonymy);
friends, friend (reiteration); and tomcat, bird, tiger (hyponymy or co-classi-
fication). The conjunctions then and when, both temporals, have replaced and,
an additive conjunction, as means of linking grammatical units. The pronouns11
she and her refer consistently within the text to Dancer, the hamster.
Children appear to have discovered the versatility of lexical ties as a
means for establishing textual relevance through synonyms, antonyms, hyponyms,
meronyms, and reiteration. Coupled with their dramatic increase in the
ability to establish identity of reference within a text, at the end of
second grade, children tied schemata together with relations of identity and
similarity. Decreases in exophoric reference indicate that they understood
that the text must be the relevant environment for all meaning relations.
Lexical cohesion, conjunction, and reference were used also to link adJaCent
propositions. And, as evidenced by the enormous increase in lexical cohesion
evidenced in their texts, children clearly grew in their ability to establish
relations at the level of categories. By the end of second grade, they had
acquired a wide range of cohesive devices and a reasonably:well organized set
of systematic options for using them. Their reference ties were routinely
endophoric. They employed substitution and ellipsis infrequently, while,
conversely, they were unhesitant in their use of conjunction. In short, so
fax as cohesion is concerned, their transition from oral to written texts,
while not complete, was well under way by the end of second grade with few
initial differences regaining among populations.
TRANSITION TO WRITING: FROM DICTATION TO WRITTEN TEXTRESULTS FOR GRADE 1-2
A major concern in our research was tog better understand how certain
features of text (e.g. cohesion) develop in stories produced orally as
compared with those produced in writing. That is, do children cope with
certain elements in oral texts, where they are free of the mechanics of
producing visual language, earlier than they do in writing? Are there
I/particular shifts in the proportional use of cohesive devices in one mode
-6- 24
that differ from those in the other? To pursue these concerns, children's
cohesive ties in dictated and written stories were compared using both
multivariati and univariate analyses. Details of the statistical analyses
and over all design strategy are described next, followed by an overview
of.the results for cohesion elements in dictation and writing.
Multivariate and Univariate Analyses
Both multivariate and univariate analyses of variance coupled with
disCriminant analysis were employed for the cohesion. comparisons. For the
multivariate and discriminant analyses, computer program CANOVA (Clyde
Computing Services, 1973) was used. This program tests for significant
differences with Wilk's likelihood ration transformed to Rao's approximate F.
SignifiCant multivariate differences were followed up with discriminant
analysts and Univariate analyses of variance. Arcsine-transformed proportions
of restricted exophoric reference, reference, substitution, conjunction, and
lexical ties served as dependent variables in a variety of complementary
multivariate analyses wherein scores for each dependent variable were well-
organized into mixed design with sex (6 males and 6 females) and dialect
(6 vernacular and-6 nonvernacular) serving as between-subjects comparisons,
and with discourse contexts (dictation and writing) and observation periods
(Spring, 1979; Fall, 1979; and Spring, 1980) serving as within-subjects
comparisons. This design compared factors within the urban school,. Similar
design arrangements were employed to compare urban and suburban schools but
with the dialect comparison removed and the school factor substituted as a
between-subjects comparison. A third multivariate analysis was employed to
compare sex (between-subjects) and mode over observations (within-subjects).
Additionally, multivariate analyses were performed on each mode separately
retaining the same design characteristics for comparisons within and between
schools. Detailed descriptions of procedures and findings are given in
Appendix A.
Essentially, the overall design strategy employee for all comps:Isons
was to consider, first, the most comprehensive model possible within the
limits of the assumption of sphericity (Davidson, 1980). Subsequent com-
plementary comparisons, premised on discriminant and univariate folloO-ups
for the comprehensive model, were designed to progress through a descending
order of less comprehensive, multivariate-univariates interaction-main-effect
probes till baseline effects were established.
Overview of Cohesion Results: Comparison of Dictated and Written Modest
Grade 1-2
Results from the urban sch::1showed that. in writing, use of conjunction
and lexical cohesion increased during grades one and two, while in dictation'',
use of conjunction was stable over the period and lexical cohesion declined
between first and second grade. In both dictation and writing, restricted
exophoric reference declined with greater decreases occurring in writing.
Both dialect groups used conjunction and lexical cohesion more in dictation
than in writing; vernacular speakers made less use of reference in writing
'than in dictation while the opposite held for nonvernacular speakers. In
general, nonvernacular speakers used more ellipeils, reference and conj4nction
in their texts overall than did vernacular speakers. The greatest differences
over the period appeared to occur in declining use of restricted exophoric
reference.
Follow-up techniques for the suburban school indicated that'differerices
over observations were associated with writing -- reference decreasing over
observations while lexical cohesion and conjunction increased over the 16-month
.period. Males employed lexical cohesion Lid ellipsis more to dictation than
26-8-
I
1
dl
1
11 in writing while the reverse was true for females. Both males and females
1
1
used conjunction less in writing than in dictation.
With the exception of restricted exophoric reference, proportions of
cohesive ties remained relatively stable over observations in the dictation
context. In writing, however, substantial developmental shifts occurred in
lexical cohesion, reference, conjunction, and ellipsis. These writing
differences were described extensively in Volume 1. Restricted exophoric
reference decreased significantly in both writing and dictation. The sta-
bility of cohesion proportions over the period in dictation suggests an
accomodation to context. On the other hand, cohesion proportions in writing
reflect significant and surprisingly swift developmental realignment to the
explicit reluiremeucs of written texts through increased lexicalization,
endophoric reference, and semantic foregrounding.
Results of MANOVA for Dictation and Writing
Retellings were excluded from these analyses OD obtain comparisons of
Children's original productions only. These productions ranged from labels
and statements, to statements with comments, through reports and personal
accounts to fictional tales. Each text, however, represented a version of
what a particular child regarded as a story. To have compared these texts
with children's retellings would have clouded interpretation of mode
differences - - assuming there is a relationship between genre and the use of
cohesive ties--for, of course, retellings consistently resulted in stories.
The fit MANOVA analyzed cohesion data from the urban school. Dialect
and'sex were between-subjects factorE and mode and observation were withir
subjects factors. Table 1 displays the means and standard deviations fo-
the cohesion proportions from the urban school by mode and observation.
-9-t
t.
27
The second MANOVA analyzed cohesion data fov the.ditation context alone
at the urban school. The between-subjects factors in this analysis remained
dialect and sex with observation the within - subjects factor. Means and
standard deviations for cohesion proportions are presented ln Table 2.
A third MANOVA analyzed cohesion data from the suburban school. Sex
was the between-subjects factor while mode and observation were within-
subjects factors. Table 18 displays the means and standard deviations for
the cohesion proportions from the suburban school.
The fourth MANOVA analyzed cohesion proportions for dictation only at
the suburban school. These means and standard deviations are available in
Table 18.
Urban School Cohesion Results: Dictation and Writin&
Results from the MANOVA on urban school cohesion proportions (see Table
3) indicated a significant multivariate test statistic for the main effects
of dialect, sex, mode, and observation. Significant interaction effects were
indicated for mode X observation and dialect X mode.
Table 3 shows a significant Wilk's lambda criterion for a first-order
interaction of mode and observation: F (10, 72) = 3.42, 2..4 .001. Note
that aftet removing effects associated with the leading root, no signifieant
discrimination remains. To determine the nature and source of this inter-
action, a discriminant analysis was performed which yielded standard
discriminant weights. Discriminant ww,ghts along with univeriate signifi-
canal tests (ANOVA) on each of the five cohesion proportions are presented
in Table 4.
Taken together, these follow-up techniques indicate that the best
discriminators for the mode by observation, first-order interaction were
conjunction, lexical cohesion and restricted exophoric reference. These
first. -order interactions are graphed in Figure 1.
-10- 28
en MIR 11111 4111. 11111P et till 111111.-71-1111.1
Table 1
Means and Standard Deviations for Cohetlion MANOVA (Transformed Variables)in Dictation and Writing at Urban School (1-2) by Mode and Observation
Mode Observation R Exo Ref Reference Ellipsis .Conjunction Lexical
M SD M SD M SD M SD M SD
Dictation 1 .07 .09 .36 .12 .01 .02 .18 .07 .40 .09
i
2 .02 .02 .45 .13 .03 .04 .20 .10 .33 .08
l-1 3 .02 .02 .47 .19 .02 .02 .18 .09 .33 .10
Writing 1 .34 .52 .31 .38 .02 .06 .04 .08 .14 .18
2 .03 .07 .35 .28 .02 .03 .10 .10 :29 .24
3 .05 .10 .39 .21 .03 .06 .15 .12 .37 .16
Observation Means 1 .21 .39 .34 .28 .01 .04 .11 .10 .27 .19
2 .02 .05 .40 .22 .02 .04 .15 .11 .31 .17
3 .04 .07 .43 .20 .02 .05 .17 .11 .36 .13
29 3U
Table 2
Means and Standard Deviations for Cohesion MANOVA (Transformed. Variables) in Dictation at the Urban School (1-2) by Dialect and Sex
Dialect Sex R Exo Ref Reference Ellipsis Conjunction Lexical
SD M SD M SD M SD M SD
Vernacular Males .07 .10 .36 .16 .01 .04 .22 .08 .35 .28
Females .03 .04 .53 .22 .02 .03 .13 .13 .33 .10
Nonvernacular Males .03 .04 .43 .05 .01 .01 .19 .07 .37 .08
Females .02 .02 .39 .08 .02 .03 .21 .05 .39 .09
31
am lior Aim as No 01._ art MI an
o Table 3
Cohesion MANOVA By Dialect, Sex, Mode,* and Observation
for Urban School (1-2)
Source df dfHYP
Between Subjects 23
Dialect (A) 1 5.00Sex (B) 1 5.00Dialect X Sex (AB) 1 5.00
S/AB 20
Within Subjects 120
Mode (C) 1 5.00Dialect X Mode (AC) 1 5.00Sex X Mode (BC) 1 5.00Dialect X Sex X Mode (ABC) 1 5.00
scIAB 20
Observation (D) 2 10.004.00
Dialect X Observation (AD) 2 10.004.00
Sex X Observation (BD) 2 10.00
4.00
Dialect X Sex X 2 10.00Observation (ABD) 4.00
SD/AB 40
Mode X Observation (CD) 2 10.004.00
Dialect X Mode X 2 10.00Observation (ACD) 4.00
Sex X Mode X 2 10.00Observation (BCD) 4.00
Dialect X Sex X Mode X 2 10.00Observation (ABCD) 4.00
SCD/AB
TOTAL 143
* Dictation and Writing Modes Only
-13-
dfERR
16.00 3.95 .016
16.00 3.01 .042
16.00 1.32 .304
16.00 12.66 .001
16.00 6.01 .003
16.00 2.17 .109
16.00 .88 .515
72.00 3.35 .001
36.50 1.01 .414
72.00 .65 .764
36.50 .61 .655
72.00 1.54 .144
36.50 .59 .674
72.00 1.20 .307
36.50 .21 .933
72.00 3.42 .001
36.50 1.50 .223
72.00 .35 .964
36.50 .35 .842
72.00 .60 .812
36.50 .15 .960
72.00 .67 .753
36.50 .48 .749
32
.35
.30
.25
.20
.15
Table 4
Discriminant Analysis and Univariate ANOVAs as Follow-up toSignificant Mode by Observation Interaction in Cohesion MANOVA
for Urban School (1-2)
Standard DiscriminantCohesion Category Function Coefficients
Univariate F Tests(2, 40)
Restricted Exophoric Reference .606 5.49 .008
Reference .330 .20 .821
Ellipsis .260 .80 .46
Conjunction - .171 5.12 .011
Lexical Cohesion . - .703 10.27 .001
R EXO REF
.10 01 03
.0502, 210mg===::::
02.01
.40
.35
.30
.25
.2
.15
.1
.05
.01
02
CONJUNCTION
01,03
.4' 01
.358/
.3
.25
Dictation Writing Dictation Writing
OBSERVATION
.15
.1
.05
LEXICAL
Dictation Writing
Figure 1. Observation as a Function of Mode for Restricted ExophoricReference, Conjunction, and Lexical Cohesion at Urban School (1-2)
-14-
33
1.
1
Mean differences (see Table 1), patterned as shown in Figure 1, indicate
that in dictation, conjunction proportions were stable over observations
(.18, .20,-.18) while small decreases occurred in lexical cohesion (.40, .33,
.35) and in restricted exophoric reference (.07, .02, 02). But in writing,
large increases occurred in conjunction proportions (.04; .10, .15) and in
lexical cohesion proportions (.14, .29, .37). Restricted exophoric reference
proportions dropped steeply in writing (.34, .03,.05). As suggested by
the standard discriminant function coefficients in Table 4, lexical cohesion
proportions and restricted exophoric reference proportions appear to contrib-
ute most substantially to the interaction; however, additional follow:gip
analyses indicated a large contribution from conjunction as well. Interpre-
tations based on standard discriminant function coefficients'are'somewhat
risky because of the possible suppression. Separate follow-up multivariate
analyses (HANOVA) of cohesion proportions in dictation, to be presented
shortly, and in writing indicated a larger role for conjunction in writing
but little contribution to the discrimination among observations_isea(ctation.
Observation and mode factor follow-ups. The significant multivariate
test statistic for the observation factor (see Table 3)1 F (10, 72) = 3.35,
It. (.001, will be examined. As shown in Table 5, the best discriminators
for observation differences, in order of decreasing importance were conjunc-
tion, restricted exophoric reference and lexical cohesion. Considered in
terms of the significant multivariate test statistic for mode (see Table 3):
F (5, 16) = 12.66,2..4 .001, in order of decreasing contribution, the best
discriminators for mode, as shown in Table 6, were: conjunction, lexical
cohesion and restricted exophoric reference. Thus, while conjunction contrib-
uted less to the interaction between mode and observation, its contribution
to both the discrimination for observation and for mode was large. Mean
Table 5
Discriminant Analysis and Univariate ANOVAs on Use of CohesionCategories in Dictation and Writing for Observation at Urban School (1-2)
Standard DiscriminantCohesion Category Function Coefficients
Univariate F Tests(2, 40)
Restricted Exophoric Reference .529 10.66 .001
Reference -.307 2.12 .13
Ellipsis -.265 .66 .52
Conjunction -.568 4.96 .01
)
Lexical Cohesion -.385 4.40 .02
differences in Table 1 for the observation factor show an increase in con-,
ction proportions (.11, .15, .17), and lexical cohesion proportions.j
(.27, .31, .36) and a decrease in proportions of restricted exophoric reference
(.21, .02; .04). Differences between modes were: conjunction (dictation:
.19; writing: .10), lexical cohesion (dictation:. .36; writing: .26) and
restricted exophoric reference (dictation: .04; writing: .14).
Dialect by mode interaction and dialect factor follow-up. As indicated .
in Table 3, a significant multivariate test statistic was obtained for the
dialect by mode interaction: F (5, 16) = 6.012..4 .003. Again, to determine
the nature and source of this interaction, discriminant analysis and uni-
variate analyses variance (ANOVA) were performed on this interaction. Standard
discriminant functipn coefficients along with the results of the univariate
significance tests on each of the five cohesion categories are given in Table 7.
These follow-up techniques demonstrate that reference and ellipsis are strong
contributors to the interaction, graphed in Figure 2 from a univariate perspective.
Table 6
Discriminant Analysis and Univariate ANOVAs on Use ofCohesion Categories in Dictation and Writing for Made at Urban School (. -2)
Cohesion CategoryStandard DiscriminatFunction Coefficients
Univariate F Tests(1, 20)
Restricted Exophoric Reference .4b6 8.54 .008
Reference -.097 3.40 .080
Ellipsis .225 .53 .48
Conjunction -.742 33.93 .001
Lexical. Cohesion -.519 29.34 .001
Table 7
Discriminant Analysis and Univariate ANOVAs as Follow-up toSignificant Dialect by Mode Interaction in Cohesion MANOVA
for the Urban School (1-2)
Standard DiscriminantCohesion Category Function Coefficients
Univariate F Tests
Restricted Exophoric Reference .286
Reference -.542
Ellipsis -.761
v
Conjunction -.453
Lexical .719
(1, 20) E. <
.79 .39
10.31 .004
10.17 .005
1.31 .27
2.75 .11
-17-36
5
.4
.30
.2
.10
REFERENCE
Vern. Nonvern.
.05-
.03-
02.01-
.00-
ELLIPSIS
Figure 2. Dictation and Writing as a Function of Dialect
at Urban School (1-2).
Mean difference (See Table 8) show that both dialect groups used
reference about equally in dictation but the vernacular group made less use
of reference than the nonvernacular group in writing. A similar pattern
held for ellipsis. Both groups employed ellipsis equally in dictation
but the vernacular group made less use of ellipsis than the nonvernacular
group in writing. However, as indicated in Table7, lexical cohesion is
contributing as well to this ftinction. The nature of its contribution to
the interaction can be inferred from the dialect and mode factor follow --ups.
.Again, using the sane combination of follow-up techniques, discriminant
and uniRariate analyses of significant multivariate test statistics, it can
\- '
be seen (Table 3) that'a significant Wilk's lambda criterion was obtained
for the dialect factor: P (5, 16) 3.95, p.' .016. Standard discriminant
function coefficients and univariate significance tests (See Table 9)
showdd that conjunction, reference and ellipsis contributed most strongly to
the dialect factor discriminant function while lexical cohesion contributed
little. Mean differences (See Table 8) show that the nonvernatular group';
-18-
37
1
SW 111111 int Eli NI Nil lei- or Nis vis
Table 84'
Means and Standard Deviations for Cohesion MANOVA (TransformedVariables) at the Urban School (1-2) by Dialect and Mode
Dialect Mode R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD M SD M SD
Vernacular .12 .28 .34 .24 .01 .03 .12 . .12 .31 ..19
Dictation .05 .07 .45 .21 .02 .03 .18 .10 .34 .10
Writing .18 .38 .24 .23 .00 .01 .07 .11' .28 .24
Nonvernacular .06 .211. .43 .23 .03 .05 .16 .09 .32 ..15
Dictation .02 .03 .41 .07 .02 .02 .20 .06 * .38 .08
Writing .10 .28 .46 .31 .04 .07 .13 .10 .25 .18
Mode Meads: Dictation .04 .06 .43 .15 .02 .03 .19 .09 .36 .09
Writing .14 .33 .35 .29. .02 .05 .10 .11 .26 .21
3833
proportions (conjunction: .16, reference: .43, ellipsis: .03) were higher
than the vernacular group's proportions (conjunction: .12, reference: .34,
ellipsis: .01), while lexical cohesion proportions were, nearly identical
(nonvernaculir: .32; vernacular: .31). Recalling the mode follow-up
(See Table 6), lexical cohesion contributed strongly to the mode discrimi-
nation, but ellipsis and reference contributed little to it. Thus, the
dialect by mode function was interpreted as having a dialect component made
up of reference, ellipsis and conjunction and a mode component comprised
of lexical cohesion, and to a much lesser extent, restricted exophoric
reference.
Table 9
Discriminant Analysie and Univariate ANOVAs on Use of CohesionCategories in Dictation and Writing for Dialect at Urban School (1-2)
Cohesion CategoryStandard Discriminaiit
Funttion Coefficients
Univariate F Tests(1, 20) 2. <
Restricted Exophoric Reference .275 2.44 .3-
Referencp -.640 4.93 .04
,..0
Ellipsis -.626 5.51 :03
Conjuiction -.748 4.86 .04
Lexical Cohesion -.004 .07 ,79
Sex factor follow-up. As indicated in Table 3, a significant multi-
variatatet statistic was obtained for sex: F (5, 16) = 3.01, 2.< .042.
The usual follow-up procedures were performed on the sex factor resulting in
a finding that lexical cohesion proportions, the only ones to achieve sig-.
nificance on the univariate test statistic, made the strongest contribution
to the sex factor discrimination (See Table 10), while conjunction, ellipsis
40-20-
1
Table 10
Ascriminant Analysis and Univariate ANO: on Use ofCohesion Categories in Dictation and Writing for Sex at Urban School (1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficient
Univariate F Tests(1, 20) <
Restricted Exophorlc Reference .324 3.50 .08
Reference -.103 0.35 .56
Ellipsis -.463 3.70 .07
Conjunction .538 3.30 .08
Lexical Cohen )n -.817 4.85 .04
and restricted exophoric, though failing to achieve significance on the
univariate statistic, made successively weaker contributions to the function.
Means and standard deviations for the sex factor are displayed in Table 11.
Girls employed a larger proportion of lexical cohesion (.35) than boys (.29).
Table 11
Means and Standard Deviations for Cohesion MANOVA (Transformed
Variables) at the Urban School (1-2) by Sex
Sex R Exo Ref Reference Ellipsis Conjunction % Lexical
M SD M SD M SD M SD M SD
Male .12 .32 .37 .26 .01 .04 .16 .11 .29 .17
Female .06 .12 .40 .21 .02 .05 .13 .09 .35 .16
Urban School Cohesion Results: Dictation. Only
The second MANOVA analyzed only the cohesion data in dictation at the
urban school. As indicated in Table 12, significant multivariate effects
were obtained for the dialect by sex interaction and the observation factor.
-21-
I.- 41
,
After removing the effects for the leading root for the observatio4.
factor, no significant discrimination remained.
Table 12
Cohesion MANOVA in Dictation, by Dialect, Sex,and Observation, for Urban School (1-2)
Source df dfHYP dfERR F E.{
Between Subjects 23
Dialect (A) 1 5.00 16.00 2.75 .056
Sex (B) 1 5.60 16.00 2.69 .060
Dialect X Sex (AB) 1 5.00 16.00 3.46 .026
S/AB 20
Within Subjects 48
Observation (C) 2 10.00 72.00 3.35 .001
4.00 36.50 1.14 .353
Dialect X Observation 2 10.00 72.00 1.43' .187
(AC) 4.00 36.50 .45 .769
Sex X Observation 2 10.00 72.00 1.53 .148
(BC) 4.00 .36.50 .63 .647
Dialect X Sex X 2 10.00 72.00 1.88 .062
Observation (ABC) 4.00 36.50 1.31 .285
SC/AB 40
TOTAL 71
Dialect by sex interaction follow-up, Again, a combination of dis-
criminant analysis and univariate test statistics were employed to probe, the
nature of significant multivariate effects. Standard discriminant function
coefficients and univariate statistics for each dependent variable are displayed
in Table 13. Lexical cohesion, reference and conjunction appear to contribute
About equally to the discrimination although lexical cohesion failed to achieve
a significant univariate test statistic. As shown by the structure coefficients
42-22-
in Table 13, reference and conjunction together, rather than combined with
lexical cohesion, appeared to be the major contributors to the function. In
all probability, suppression accounted for the discrepancy between these two
coefficients. Structure coefficients represent correlations between discrimi?
nant scores and the original variables; these correlations provide estimates
of the function with suppression removed. Thus, the interaction was inter-
preted as one largely based upon the contributions of reference and conjunction.
Table 13
Discriminant Analysis and Univariate ANOVAs as Follow-up toSignificar..z Dialect by Sex Interaction in Cohesion MANOVA
in Dictation for Urban School (1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Structure Univariate F TestsCoefficients (1, 20) 2.(
Restricted Exo Ref 1.178 .450 2.54 .13
Reference 2.430 -.829 9.17 .007
Ellipsis 1.043 -.157 .29 .60
Conjunction 2.384 .821 10.02 .005
Lexical Cohesion 2.541 .199 .63 .44
Figure 3 graphs these first-order interactions from a univariate perspective.
Vernacular females made greater use of reference and less use of conjunction
than the other groups; nonvernacular females followed the opposite pattern- -
using reference less that nonvernacular males but conjunction more. Reference
and conjunction means for nonvernacular females were similar to those for
vernacular males (See Table 2).
-23- 43
.50
REFERENCE
vtates
Vern. Nonvern.
. 50i
.40-
.30-
.20-
.10-
CONJUNCTION
4Vern. Nonvern.
Figure 3. Reference and Conjunction as a Function of Dialect in Dictationat Urban School (1-2)
Observation factor follow -up. The significant multivariate test'
statistic for the observation factor: F (10, 72) 3.35,2.4 .001, was
followed up by discriminant and univariate analyses (See Table 14).
Table 14
Discriminant Analysis and Univariate ANOVAs on Use of CohesionCategories in Dictation for Observation at the Urban School (1-2)
Cohesion Category
Standard DiscriminantFunction Coefficients
Univariate F Tests(2, 40)
Restricted Exophoric Reference .745 6.56 .003
Reference .166 5.17 .01
Ellipsis -.394 1.98 .15
Conjunction .255 .27 .77
Lexical Cohesion .846 7.76 .001
Lexical cohesion and restricted exophoric reference appeared to make the
strongest contribution to the function from a multivariate perspective,
while from a univariate framework, reference appeared also to account for
significant observation differences in cohesion proportions. Mean differ-
ences over observation (See Table 15) show a drop in exophoric reference
and lexical cohesion at observation 2 (early second grade) accompanied
by an increase in reference. Cohesion proportions in all categories for
dictation remained fairly stable throughout grade two.
Table 15
Means and Standard 'Deviations for Cohesion MANOVA (TransformedVariables) in Dictation at Urban School (1 -2) by Observation
Observation R Exo Ref Ref. Ellips Conj. Lex.
1 M .07 .36 .01 .18 .40
SD (.09) (.12) (.02) (.07) (.09)
2 M .02 .45 .03 .20 .33
SD (.02) (.12) (.04) (.10) (.08)
3 M .02 .47 .02 .18 .35
SD (.02) (.19) (.02) (.09) (.10)
Suburban School Cohesion Results: Dictation and Writing
The third MANOVA analyzed cohesion data from subjects at the suburban
school. Results from this multivariate analysis of variance are presented
in Table 16. The MANOVA summary table indicates significant first-order
interactions for sex by mode: F (5, 6) = 8.55, 2.4 .01 and mode by
observation: F (10, 32) = 3.82, 2...4 .002, as well as a main effect for
the observation factor: F (10, 32) = 3.33, II.< .005.
Table 16
Cohesion MANOVA by Sex, Mode, and Observationin Dictation and Writing for Suburban School (1-2)
Source df dfHYP dfERR F 2.
Between Subjects 11
Sex (A) 1 5.00 6.00 1.57 .30
S/A 10
Within Subjects 60
Mode (B) 1 5.00 6.00 1.67 .27
Sex X Mode (AB) 1 5.00 6.00 8.55 .01
SB/A 10
Observation (C) 2 10.00 32.00 3.33 .005
4.00 16.50 1.31 .31
Sex X Observation (AC) 2 10.00 32.00 .77 .65
'COO 16.50 .79 .55
SC/A. 20
Mode X Observation (BC) 2 10.00 32.00 3.82 .002
4.00 16.50 1.03 .42
Sex X Mode X 2 10.00 32.00 1.83 .09
Observation (ABC) 4.00 16.50 1.71 .20
SBC/A 20
TOTAL 71r
Mode by observation interaction and observation follow-ups. Discriminant
analysis and univariate tests of significance (See Table 17) indicated that
the interaction function was made up largely of lexical cohesion with lesser
contributions from ellipsis, exophoric reference and reference. However,
only lexical cohesion and reference achieved significance from a univariate
perspective. Lexical cohesion and reference are graphed in Figure 4.
11
IFIF
Table 17
Discriminant Analysis and Univariate ANOVAs on Use of Cohesion Categoriesin Dictation and Writing for Mode by Observation Interaction
at Suburban School (1-2)
Standard DiscriminantCohesion Category Function Coeffic ..ents
Univikiate P Tests(2, Z0) E.
Restricted Exophoric Reference 0.158 0.06 .94
Reference
Ellipsis
-0.068
-0.333
4,
6.02
0.84
.009
.45
Conjunction -0.029 0.18 .84
Lexical Cohesion -1.065 25.10 .001
. 65
.60-
. 55
.50-
.45-
.40-
. 35-
. 30-
.25-
.20-
. 15-
. 10-
.05-
.01-
01
03
02
03
02
01
f
.65-
.60-
.55-
.50-
.45-
.4 0-
.35-
.30-
.25-
.20-
.15-,10-.05-.01-
030 1,0 2
01
Dictation Writing Dictation Writing
LEXICAL COHESION REFERENCE
Figure 4. Lexical Cohesion and Reference in Dictation and Writing
as a Function of Observation (1-2).
Mean difference (See Table 18) show that lexical cohesion proportions increased
at each observation in writing while they declined in dictation early in gkade
two and rose only slightly by the end of second grade. Use of reference was
stable across dictation observations but dropped across observations for writing.
-27- 47
t.,
CO
Table 18
Means and Standard Deviations for Cohesion MANOVA(Transformed Variables) at the Suburban School (1-2) by Mode and Observation
Mode Observatioft R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD .M SD M SD
Dictation 1 .03 .04 .39 .10 .01 .01 .16 .08 .45 .11
2 .02 .02 .39 .08 .02 .02 .25 .06 .34 .09
3 .01 .01 .41 .08 .01 .01 .20 .07 .39 .06
Writing 1 .03 .10 .65 .32 .01 .02 .11 .13 .29 .20
2 .02 .03 .41 .12 .03 .03 .17 .09 739 .11
3 .01 .02 .36 .07 .01 .02 .15 .06 .50 .11
Observation- Means 1 .03 .03 .52 .27 .01 .01 .13 .11 .37 .18
2 .02 .03 .40 .10 .02 .03 .21 .08 .37 .10
3 .01 .01 .38 .08 .01 .02 .17 .07 .47 .10
= 48 49
.
so um op eel Nor me am is am nap lie al me on me an gm
Follow-up discriminant analysis and univariate tests of significance
for the observation factor (See Table 19) showed that reference was the
largest contributor to the function followed by lexical cohesion, ellipsis,
and conjunction. Mean differences for observations (See Table 18) indicated
an overall increase in lexical cohesions --as shown above largely in writing- -
and a decrease in reference -- again, largely in writing. :21ipsis and
conjunction proportions rose at observation two (early grade two) but fell
slightly at observation three (end of grade two).
Table 19
Discriminant Analysis and Univariate ANOVAs on Use of CohesionCategories in Dictation and Writing for Observation at Suburban School (1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Univariate F Tests(2, 20)
Restricted Exophoric Reference 0.095 1.15 .34
Reference 1.205 6.8$ .005
Ellipsis -0.598 5.97 .009
Conjunction -0.183 7.47 .004
Lexical Cohesion .722 2.94 .08
Sex b mode interaction follow-u in dictation and writin:. Discriminant
and univariate analyses were employed to follow up the sex by mode interaction:
F (5, 6) me 8.55, 2. < .01. All cohesion categories appear to have contributed
to the discrimination with only ellipsis achieving significance on the uni-
variate tests of significance (See Table 20). As will be shown in the next
analyses for dictation alone, and can be seen from the previous analyses of
writing alone (King 6 Rentel, 1981, pp. 60-61), no differences were obtained
for the sex factor in dictation but, in writing, ellipsis and lexical
cohesion were significantly different for the sex factor -- ellipsis: F (1, 10)
= 30.83, 2..< .001; lexical cohesion! F (1, 10) = 5.23, R.< .05.
Table 20
Discriminant Analysis and Univariate ANOVAs on Use of CohesionCategories in Dictation and Writing for Sex by Mode
Interaction at Suburban School (1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Univariate F Tests'(1, 10) 2.4
Restricted Exophoric Reference 1.188 1.16 .31
Reference 1.736 1.01 .34
Ellipsis -0.928 37.09 .001
Conjunction 0.638 2.57 .14
Lexical Cohesion .644 3.86 .08
As shown in Figure 5, male and female ellipsis proportions were equal in
dictation but female proportions were higher in writing. Means and standard
deviations are displayed in Table 21.
.50
.45
.40
.35
.30
.25-
.20-
.15-
.10-
.05-
.01-
ELLIPSIS
Males
es
Dictation Writ ng
Figure 5. Ellipsis as a Function of Sex in Dictation and Writing
at Suburban School (1-2).
51-30-
.
NM Ile NI MS or el es am no MI me MI ma
Table 21
Means and Standard Deviations for Cohesion MANOVA (TransformedVariables) by Sex and Mode in Dictation and Writing at Suburban School (1-2)
Sex Mode R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD M SD ?I SD
MalesDictation .02 .03 .40 .10 .01 .02 .21 .09 .38 .11'
Writing .03 .08 .52 .31 .00 .01 .18 .10 .33 .19
Females1Dictation .02 .02 .39 .07 .01 .01 .19 .06 .41 .09
.
Writing .00 .01 .43 .13 .03 .03 .11 .09 .46 .12
53
Suburban School Cohesion Results: Dictation Only
The fourth MANOVA analyzed only cohesion proportions in dictation at
the sub urban school. A significant multivariate test statistic was obtained
only on the observation factor for the leading root with a marginal statistic
for the second root (See Table 22). Only the leading,root was interpreted;
however, both are presented in Table 23, where discriminant and univariate
follow-ups are displayed. Lexical cohesion appeared to make the strongest
Table 22
Cohesion MANOVA in Dictation by Sex and Observationfor Suburban School. (1 -2)
Source df dfHYP dfERROMMIF Z. <
Between Subjects 11
Sex (A) 1 5.00 6.00 1.18 .42 t
STA 10
Within Subjects 24
-Observation (B) 2 10.00 32.00' 2.15 .05
4:00 16.50 2.62 .07
Sex X Observation 2 10.00 32.00 1.23 .31
(AB) 4.00 16.50 .19 .94
SBA 20
TOTAL 35
contribution to the function (first root) followed by reference and conjunc-
tion. However, since reference did not achieve a significant univariate test
statistic, and given its strong correlation with lexical Cohesion (.62),
suppression was suspected; thus, only significant univariate differences for
lexical cohesion and conjunction were interpreted. Means and standard
deviations were presented in Table 18. Summarized, for lexical cohesion, .
Tab-a 23
Discriminant Analysis and Univariate ANOVAs in Dictationfor Observation Factor at Suburban School (1-2)
Cohesion Category
Standard DiscriminantFunction CoefficientRoot Root1 2
Restricted Exophoric Reference 5.064
Reference 12.166
Ellipsis 1.790
Conjunction 10.203
Lexical Cohesion 12.440
Univariate F Tests(2, 20) 2. 4
.548 1.47 .25
.607 .20 .82
-.470 3.23 .06
-.068 4.46 .03
.731 4.12 .03
Proportions were: .45, .34, and .39 across observations; for conjunction,
proportions were: .16, .25, sad .20 across observations. Use of lexical
cohesion dropped at observation two and increased at observation three.
Conversely, use of conjunction increased at observation two and dropped at
observation three.
Discussion of these results begins on page 45 following the presentation
of results for the kindergarten/grade 1-2 replication study which follows.
5-33-
4
1
Chapter 2
Kindergarten/Grade One Replication of the Transition from
Speech to Writing
Our search for factors influencing children's transition from speech to
writing began with subjects mid-way through first grade. Before long, however,
we recognized that our subject (6 and 7 year olds) may already have passed
through some of the critical factors in development. Therefore, we selected
a comparable sample of kindergarten children in the spring of the school
year to observe and to contrast their development with that of the older
populati8ii. This additional study provided not only opportunity to probe for
those factors we assumed occurred earlier in development, but also increased
the potential generaizability of the research findings. Observations, data
collection procedures and analyses of data for the kindergarten-grade one
0 -1) subjects were identical to those used for the older children (See Appendix
A for Procedures).
The main question prompting our research concerned the relationship
lAtween children's oral modes (story retelling and dictation) and 'their
Witten mode. Srscifically, are there features in stories produced orally
that foreshadow charateristics in their writing? Are there trends in
development in the different modes that are similar? Also, when do children
from kindergarten through grade two learn to cope with particular demands
of written texii7--This chapter aeeresses these queutiwis-by-presenting-coheeion
results of the replication study (K-1/1-2) for dictation and writing only:
As noted earlier, it was not feasible to compare these modes win story re-
tailing for the total population. Chapter III gives results of story structure
comparisons across the three modes for those children who wrote and dictated
-34-
I/
stories from the beginning. The results which follow compare K-I to Grade
1-2 children on the basis of their use of elements of cohesion in dictated
and written stories.
COHESION RESULTS FOR DICTATION AND WRITING:REPLICATIONS (K-1/1-2)
Preliminary multivariate analyses for school and mode indicated
significant test statistics for both school and mode factors (See Appe4dix B).
These findings led to a decision to compare kindergarten and grade one popu-
lations within each school in dictation and in writing separately. Thus,
four multivariate analyses of variance (MANOVA) were performed on arcsine-
transformed proportions of the five cohesion categories noted earlier,
The first MANOVA analyzed cohesion proportions in dictation at the urban
school were grade (K1/1-2), dialect (vernacular-nonvernacular), and sex
(male-female) served as between-subjects factors and observation (March, 1979;
October, 1979; and May, 1980) served t- a within-subjects factors (See Table
24). Identical design arrangements were employed in the second MANOVA to
analyze cohesion proportions in writing at the urban school. Means and
standard deviations from the dictation MANOVA are presented in Table 25. 1The third MANOVA analyzed cohesion proportions in dictation from the
suburban school. Grade (K-1/1-2) and sex (male-female) ware between-subjects
factors; observation (March, 1979; October, 1979; and May, 1980) was a within-
subjects factor. The sane design arrangements were employed in the final
MANOVA to analyze cohesion proportions in writing.
11
IUrban School Cohesion Results for Dictation: Replication (K- 1/1 -2)
11
Results from the cohesion MANOVA on the urban-school, grade-replication
data in dictation (See Table 24) indicated a significant multivariate test
statistic for the observation factor. None of the first- or second-order
interations were significant. Table 25 shows a significant Wilk's lambda
-35- 57
1
Table 24
Cohesion MANOVA, by Grade, Dialect, Sex, and Observationin Dictation for Urban School (K-1/I-2)
Source df OfF1T dfERR F
Between Sublects 39
Grade (A) 1 5.00 28.00 .88 .51
Dialect (B) I 5.00 28.00 .49 .79
Sex (C) 1 5.00 28.00 1.00 .44
Grade X Dialect (AB) 1. 5.00 28.00 1.01 .43
Grade X Sex (AC) 1 5.00 28.00 2.08 .10
Dialect X Sex (BC) I 5.40 28.00 .72 .62
Grade X Dialect X Sex (ABC) 1 5.00 28.00 1.88 .13
S/ABC 32
Within Subjects 80
Observation (D) 2 10.00 120.00 4.51 .001
4.00 60.50 4.70 .002
Grade V Observation (AD) 2 10.00 120.00 1.74 .08
4.00 60.50 .86 .49
Dialect X Observation (BD) 2 10.00 120.00 .51 .88
4.00 60.50 .21 .93
Sex X Observation (CD) 2 10.00 120.00 1:14 .34
4.00 60.50 .18 .95
Grade X Dialect X Observation 2 10.00 120.00 1.17 .32
(ABD) 4.00 60.50 .29 .89
Grade X Sex X Observation 2 10.00 120.00 .61 .80
(ACD) 4.00 60.50 .51 .73
Dialect X Sex X Observation 2 1G.00 120.00 .93 .51
(BCD) 4.00 60.50 .77 .55
Grade_X_Dialest_k_s_tx IL 2 10.00 120.00 1.00 .45
Observation (ABCD) 4.00 60.50 .26
SD;ABC 64
TOTAL 119
-36-
58
I
Table 25
Means and Standard Deviations of Cohesion MANOVA (Tranpforued Variables)in Dictation for Urban School (K-1/1-2) -- by Grade and Observation
Grade Observation R Exo Ref Reference Ellipsis Conjunction Lexical
SD M SD
K-1
1-2
Observation Means
59
M
.05
SD
.08
M
.42
SD M
.02
SD
.03
M
.17.13
1 .10 .12 .42 .11 .02 .04 .15
2 .03 .04 .42 .14 .02 .02 .21
3 .03 .04 .40 .16 .00 .00 .14
.04 .06 .43 .15 .02 .03 .19
1 .02 .03 .36 .12 .01 .02 .18
2 .02 .03 .45 .13 .03 .04 .20
3 .02 .02 .k7 .19 .02 .02 .18
1 .08 .10 .39 .12 .01 .03 .17
2 .03 .03 .43 .12 .03 .03 .21
3 .02 .03 .42 .75 .01 .02 .17
.10 .38 .17
.09 .32 .09,
.12 .33 .07
.08 .47 .24
.09 .36 .09
.07 .40 .09
.10 .33 .08
.09 .35 .10
.08 .36 .10
.11 .33 .07
.08 .42 .19
60
as sa aim NM ON MO OW Mlle MI OM NIP OM MN 41111 AID OS
criterion for observation for the first root: P (10, 120) = 4.51, E.< .001;
and the second root: F (4.00, 60.50) = 4.70, E.< .002.
Observation toctor follow-up. Discriminant analysis and univariate
analyses of variance (ANOVA) were employed to probe the nature and source of
the significant, multivariate, test statistic for the observation factor.
These follow-up techniques (See Table 26) indicated that, lexical cohesion
and conjunction contributed about equally to the function making up the first
root, with ellipsis and restricted exophoric references contributing lesser
Table 26
Standard Discriminant Function Coefficients and Univariate ANOVASon Use of Cohesion Categories in Dictation for Observation Factor at
Urban School (Itt/1-2)
Standard DiscriminantFunction Coefficients Univariate F Tests
Cohesion Category Root Root (2, 64) R.<1 2
Restricted Exophoric Reference -0.925 -0.507 9.43 .001
Reference -3.213 -0.133 1.02 .37
Ellipsis -1.024 -0.535 3.59 .03
Conjunction -2.474 -0.417 2.77 .07
Lexical Cohesion -2.407 0.507 5.68 .005
amounts to the discrimination, while ellipsis, lexical cohesion, restricted
exophoric reference and conjunction contributed about equally to the function
making up the second root. Mean differences (See Table 25) indicated overall
decreases in restricted exophoric reference with increases in cohesion propor-
tions for other categories across observations--reference excepted, given its
relative stability over observations. This stability probably accounts for
its strong contribution to the function making up the first root. The
presence of two significant roots, composed of similar contributions but
distinguished by, differences in directionality for lexical cohesion between
first and second roots, is probably best explained by increasing lexical
cohesion proportions across observations for the K -i population contrasted
with decreasing lexical cohesion proportions for the 1-2 population. This
explanation is supported by the trade by observation interactionwhich
approached a significant multivariate test statistic (See Table 25) and,
indeed achieved a designated level on the univariate test statistic only
for lexical cohesion: F (2, 64) im 6.02, p. < .001 (See Table 27). This
univariate grade by observation interaction for lexical cohesion is graphed
in Figure 6
.50-
.45-
.40-
.35-
.30-
.25-
.20-
.15-
.10-
.05-
.01
03
01
020201
K-I 1-2
LEXICAL COHESION
Figure 6. Observation as a Function of Grade inDictation at Urban School (K-1/1-2).
-39-62
IIIIIII
IIIIII1
iI
I1
II
Table 27
Discriminant Function Coefficients and Univariate ANOVAs on Use ofCohesion Categories in Dictation for Grade by Observation
Interaction at Urban School (K-1/1-2)
Cohesion Category.
Standard DiscriminantFunction Coefficients
Univariate F Tests(2, 64) 2..c
Restricted Exopboric Reference .627 .33 .72
Reference .972 1.46 .24
Ellipsis .665 1.68 .19
Conjunction .723 1.65 .20
Lexical Cohesion -.080 6.02 .004
T
Urban School Cohesion Results for Writing: Replication (K-111-2)
Results from the MANOVA on the urban-school, K-1/1-2 data indicated
a significant multivariate test statistic for the main effects of grade,
dialect, and observation; and first-order interaction effects for grade
by dialect and grade by sex (See Table 28).
Grade by dialect interaction in writing at urban school. Using the
customary discriminant analysis and univariate analysis of variance (ANOVA)
procedures to follow up significant multivariate effects, these techniques
indicated lexical cohesion proportions contributed strongly to the discrim-
ination followed by reference, conjunction and ellipsis. Only lexical
cohesion and ellipsis achieved predesignated alpha levels on the univariate
significance tests (See Table 29).
-40v 63
Table 28
Cohesion MANOVA by Grade, Dialect, Sex, and Observationin Writing at Urban School (K-1/1-2)
Source of dfHYP dfERR 2:4
Between Subjects 39
Grade (A) 1 5.00 28.00 4.43 .004
Dialect (B) 1 5.00 28.00 4.28 .005
Sex (C) 1 5.00 28.00 .86 .52
Grade X Dialect (AB) 1 5.00 28.00 7.23 .001Grade X Sex (AC) 1 5.00 28.00 3.47 .02Dialect X Sex (BC) 1 5.00 28.00 1.43 . .25
Grade X Dialect X Sex 1 5.00 28.00 .50 .78
(ABC)
S/ABC 32
Within Subjects 80
Observation (D) 10.00 120.00 5.11 .001
4.00 60.50 1.49 .22
Grade X Observation 10.00 120.00 1.62 .11
(AD) 4.00 60.50 0.94 .45
Dialect X Observation 10.00 120.00 0.57 .84
(BD) 4.00 60.50 0.07 .99
Sex X Observation (CD) 10.00 120.00 1.13 .34
4.00 60.50 0.89 .48
Grade X Dialect X 10.00 120.00 1.09 .37
Observation (ABD) 4.00 60.50 0.38 .82
Grade I Sex X 2 10.00 120.00 0.60 .81
Observation (ACD) 4.00 60.50 0.34 .85
Dialect X Sex X 2 10.00 120.00 1.05 .41
Observation (BCD) 4.00 60.50 0.68 .75
Grade X Dialect X Sex 2 10.00 120.00 0.27 .99
X Observation (ABCD) 4.00 60.50 0.23 .92
SD/ABC 64
TOTAL 119 .
-41- 64
1
Table 29
Discriminant Function Coefficients and Univariate ANOVAs on Use ofCohesion Categories in Writing for Grade byDialect Interaction at Urban School (K-1/1-2)
Standard DiscriminantCohesion Category Function Coefficients
Univariate F Tests(1, 32) 2:4
Restricted Exophoric Reference .044 .04 .84
Reference 1.100 1.44 .24
Ellipsis .458 4.53 .04
Conjunction .740 1.16 .29
Lexical Cohesion -1.400 5.64 .02
Mean differences for ellipsis and lexical cohesion proportions are graphed
in Figure 7 with corresponding means and standard deviations displayed in
.50.- .50-
.45- .45-
.40 .40-
.35- .35T
.30-N1
4ly .30-
.25- e .25-
.20- 4,e, 4 .20-t,
It
.15- v4 0,. .15-
.10-4 oo ego v.
40.05- 4 .05-.01-
-..--...------ 1,
.01-
nonvernacular
K-1 1-2 K-1 1-2
ELLIPSIS LEXICAL COHESION
Figure 7. Dialect as a Function of Grade for Ellipsis andLexical Cohesion in Writing fir Urban School (K-1/1-2)
Table 30. Kindergarten vernacular and nonvernacular ellipsis ;means were
roughly equal while 1-2 nonvernacular means for ellipsis were higher
than 1-2 vernacular means. Figure 7 also shows that nonvernacular K-1 ,
means for lexical cohesion were higher than vernacular means but that
1-2 nonvernacular and vernacular means for lexical cohesion were roughly .
equal. In other words, for lexical cohesion earlier differences between
vernacular and nonvernacular subjects all but disappeared in the grade1
1-2 population. The reverse was true for ellipsis: differ4nces were
observed in the grade 1-2 population that did not exist in be K-1
population.
Grade by sex interaction in writing at urban school (K-1/1 -2) - -follow -up.
A significant multivariate test statistic for the grade by sex interaction
(See Table 28 ), again, was followed up by a combination of discriminant
analysis and analysis of variance on'the five cohesion pioportions for
writing at the urban school. None of the univariate test statistics for
cohesion proportions achieved predesignated alpha levell, although lexical
cohesion, one of the strongest contributors to the discriminant function,
t.did approach significance: F (1, 32) = 3.39, p.< .08.; As shown in Table
31, lexical cohesion, reference, and conjunction contributed strongly to
the function; however, in the absence of significant/univariate effects,
interpretation of this grade by sex interaction was/simply too risky to
pursue. The absence of significant multivariate ox univariate main effects
fog sex coupled with significant multivariate and univariate effects for
grade suggest a guarded grade level interpretation. Means and standard
deviations by sex and grade are given in Table 32 for the reader who may
want to pursue an interpretation.
-43- 66
me ma UN so Ns ea on is so. in on ow a an um I_, al MN.
Table 30
Means and Standard Deviations for Cohesion MAMMA (TransformedVariables) on Writing for Urban School (K-1/1-2) -- by Grade and Dialect
Grade Dialect R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD , M SD M SD
K-1 .07 .24 .34 .39 .01 .02 .03 .08 .20 .24
Vernacular .11 .33 .34 .50 .00 .01 .02 .06 .12 .20
Nonvernacular .04 .09 .34 .27 .01 .02 .05 .09 .27 .26
1-2 .12 .31 .32 .26 .02 .06 .10 .11 .27 .22
g'As
Vernacular .15 .32 .23 .23 .01 .01 .07 .12 .28 .250- Nonvernacular .10 .30 .41 .25 .04 .07 .13 .10 .27 .18
Dialect Means:
Vernacular .13 .32 .29 .39 .00 .01 .04 .10 .20 .24
Nonvernacular .07 .22 .37 .26 .03 .06 .09 .11 .27 .22
67 es
Table 31
Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Writing for Grade by Sex
Interaction at Urban School (K-1/1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Univariate F Tests(1, 32) 2.* <
Restricted Exaphoric Reference .282 1.12 .30
Reference 1.169 .32 .57
Ellipsis -.062 .59 .45
Conjunction .704 1.40 .25
uexical Cohesion -1.416 3.39 .08
69
-45-
r-i 1 I IIIIIII MIII IMO IN OM OM In IMO IIIIIII IIIIII UM Mill OM
Table 32
Means and Standard Deviations for Cohesion MANOVA (Transformed Variables)on Writing for Urban School (K-1/1-2) -- by Grade and Sex
Grade Sex R Exo Ref Reference Ellipsis Conjunction Lexical
K-1
M SD M SD M SD. M SD M SD
Males .07 .13 .31 .36 .00 .01 .03 .07 .21 .25
Females .03 .26 .38 .41 .01 .01 .04 .08 .18 .19
1-2 Males .19 .44 .36 .35 .01 .04 .11 .13 .21 .20
Females .09 .16 .34 .23 .03 .06 .08 .09 .32 .21I
7t)
Grade factor follow-R. Given that dialect and sex were implicated I
with grade in previously discussed interaction effects, the following
interpretation of the grade-factor, discriminant and uivariate, follow-up
techniques should be taken with care. These follow-up techniques (See
Table 33) indicated that, of the cohesion proportions achieving significant
Table 33
Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Lriting for Grade Factor
at Urban School (K-1/1-2),
Cohesion Category
1. IStandard DiscriminantFunction Coefficie-Its
Univariate" Tests(1, 32) 2.. <
Restrtcted Exophoric Reterence -.060 .96 .34
Reference -.427 .09 .76
Ellipsis .643 7.12 .01
Conjunction .648 12.16 .001
Lexical Cohesion .442 4.99 .03
univariate alpha levels, conjunction, ellipsis and lexical cohPA4cn, in that
order, made the strongest contribution to the discriminant function. G'.:ac'es
level means and etandatd deviations for cohesion pror.irtions are given in
Table 25. All grade 1-2 proportions were higher than K-1 proportions;
conjunction (.10, .03); ellipsis (.02, .01) and lexical cohesion (.27, .20).
Dialect factor follow-up. Follow-up techniques for the significant
multivariate main effect for dialect (See Table 28) indicated that ellipsis
and conjunction were the,strongest contributors to the discriminant function
and both, os indicated in Table 34, achieved designated univariate alpha levels.
-47-7 1
Table 34
Discriminant Function Coefficients and Univariate ANOVAs on Use ofCohesion Categories in Writing for Dialect Factor at
Urban School (K-1/1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Univariate F Tests(1, 32) 4
Restricted Exophoric Reference .437 1.66 .20
Reference -.077 1.39 .25
Ellipsis -.788 11.08 .002
Conjunction -.718 5.53 .03
Lexical Cohesion .011 3.46 .07
Dialect means are displayed in Table 30. Nonvernacular means were higher than
vernacular means for both ellipsis (.03, .00) and conjunction (.09, .04),
Observation factor follow-up. Discriminant function coefficients and
univariate ANOVAs, performed on the writing data, following up the significant
multivariate statistic for observation (See Table 28), indicated that all
cohesion proportions except ellipsis, contributed to the observationdis-
crimination -- lexical cohesion making the strongest contribution, followed
by conjunction, reference, and restricted exophoric reference. (See Table 35),
Observation means for cohesion proportions in writing at the urban school are
displayed in Table 36. Lexical cohesion proportions increased across observa-
tions as did proportions for c=4,:nction and reference. Restricted exophoric
reference proportions declined as well but not uniformly.
Table 35 IDiscriminant Function Coefficients and Univariate ANOVAs on Use of
Cohesion Categories in Writing for Observation Factor atUrban School (K1/1-2)
Standard Discriminant Univariate F TestsCohesion Category Function Coefficients (2, 64) k..c.
Restricted Exophoric Reference -.136 3.94 .02
Reference -.425 3.70 .03
Ellipsis -.088 .27 .76
Conjunction -.464 8.93 .001
Lexical Cohesion -.777 * 16.39 .001
11_
II
I
II
II
ITable 36
Means and Standard Deviations of Cohesion MANOVA (Transformed Variables)on Writing for Urban School (K-1/1-2) --by Observation
Observation R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD M SD M SD
1 .19 .37 .24 .29 .01 .05 .02 .07 .11 .17
2 .01 .04 .34 .35 .02 .03 .08 .10 .23 .24
II3 .09 .28 .41 .34 .02 .05 .10 .12 .37 .21
II
Suhrban School Cohesion Results for Dictation: Replication (K-1/1-21dB
The third-MANOVA analyzed cohesion data in dictation at the suburban
school. No significant main or interaction effects were obtained for rte Ill
dictation mode, although the observation factor approached a significant
IIWilk's lambda ctiterion (See Table 37).
II
73-49-
Table 37
Cohesion MANOVA by Grade, Sex, and Observationin Dictation for the Suburban School (K-I/1-2)
Source df dfHYP dfERR F P.4
Between Subjects 23
Grade (A) 1 5.00 16.00 .68 .64
Sex (B) 1 5.00 16.00 1.10 .40
Grade X Sex (AB) 1 5.00 16.00 1.39 .28
S/AB 20
Within Subjects 48
Observation (C) 2 10.00 72.00 1.83 .06
4.00 36.50 1.74 .16
Grade X Observation (AC) 2 10.00 72.00 .80 .63
4.50 36.50 .50 .74
Sex X Observation (BC) 2 10.00 72.00 .71 .71
4.00 36.50 .29 .88
Grade X Sex X 2 10.00 72.00 .66 .76
Observation (ABC) 4.00 36.50 .68 .61
SC/AB 40
TOTAL 71
Suburban School Cohesion Results for Writing: Replication (K-1/1-2)
The final cohesion MANOVA analyzed writing data at the suburban school.
Significant multivariate test statistics were obtained for the sex and
observation factors as well as for the grade by observation and sex by ob-
servation, first-order interactions (See Table 38). Cohesion means and
standard deviations for dictation and writing by grade and,observation are
displayed in Table 39.
74
-SO-
Table 38
Cohesion MANOVA by Grade, Sex, and Observation
in Writing for the Suburban School (K-1/1-2)
Source df dflUP dfERR 2.<
Between Subjects 23
Grade (A) 1 5.00 16.00 2.14 .11
Sex (B) 1 5.00 16.00 2.95 .05
Grade X Sex (AB) 14 5.00 16.00 1.84 .16
S/AB 20
Within Subjects 48
Observation (C) 2 '0.00 72.00 4.48 .001
4.00 36.50 .99 .43
Grade X Observation (AC) 2 10.00 72.00 2.39 .02
4.00 36.50 1.21 .32
Sex X Observation (BC) 2 10.00 72.00 2.14 .03
4.00 36.50 .71 .59
Grade X Sex X 2 10.00 72.00 1.84 .07
Observation (ABC) 4.00 36.50 .92 .46
SC/AB 40
TOTAL 71
-51-
In SIM MI MI WM MI IND MI =I OM se us Ns on ErTable 39
Means and Standard Deviations of Cohesion MANOVA (Transformed Variables)
in Dictation and Writing for Suburban. School (K-1/1-2) by Grade, Mode, and Observation
Grade Mode Observation R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD M SD M SD
K-1 Dictation 1 .0'. .02 .44 .20 .01 .02 .20 .04 .36 .15
2 .01 .01 .38 .07 .01 .02 .22 .10 .40 .09
3 .00 .01 .47 .06 .02 .04 .18 .07 .35 .09
Writing 1 .04 .10 .48 .45 .01 .02 .04 .06 .16 .17
2 .07 .21 .45 .28 .14 .45 .09 .12 .36 .21
3 .00 .00 .47 .19 .03 .06 .13 .11 .40 .13
1-2 Dictation 1 .03 .04 .39 .10 .01 .01 .16 .08 .45 .11
2 .02 .02 .39 .08 .02 .02 .25 .06 .34 .09
3 .01 .01 .41 .08 .01 .01 .20 .07 .39 .06
Writing 1 .03 .10 .65 .32 .00 .02 .11 .13 .29 .20
2 .02 .03 .41 .12 .03 .03 .17 .09 .39 .11
3 .01 .02 .36 .07 .01 .02 .15 .06 .50 .11
Means OverGrade:
Dictation .01 .02 .41 .11 .01 .02 .20 .07 .38 .11'
1 .02 .03 .41 .1: .01 .02 .18 .06 .41 .14
2 .02 .02 .39 .07 .02 .02 .23 .08 .37 .10
3 .00 .01 .44 .07 .02 .03 .19 .07 .37 .08
Writing .03 .10 .47 .28 .04 .19 .12 .11 .35 .19
1 .03 .10 .56 .40 .01 .02 .08 .11 .23 .20
2 .04 .15 .43 .21 .09 .32 .13 .11 .38 .17
3 .00 .01 .42 .16 .02 .06 .14 .09 .45 .13
b0" 77
Grade by observation follow up. To determine the nature of the sig-
nificant multivariate test statistic for the grade by observation interaction
in writing at the suburban school, as in previous follow-ups, discriminant
and univariate analyses were employed (See Table 40). Reference, ellipsis,
Table 40
Standard Discriminant Function Coefficients and Univariate ANOVAson Use of Cohesion Categories in Writing for Grade by
Observation at Suburban School (K-1/1-2)
Cohesion CategoryStandard DiscriminantFunction Coefficients
Univariate F Tests(2, 40) 2.<
Restricted Exophoric Reference .664 .46 .64
Reference 1.421 1.87 .17
Ellipsis 1.092 .60 .56
Conjunction .828 .72 .49
Lexical Cohesion .908 .85 .44
lexical cohesion, conjunction, and restricted exophoric referedce, in that
order, contributed successively lesser amounts to the discrimination. No
significant, univariate, first-order interactions were obtained far any of
the cohesion proportions. Examination of the writing means (Table 39) for
each grade across observations are graphed in Figure 8. Proportions of
reference, conjunction and lexical cohesion are generally higher for the
1-2 population and somewhat higher for the K-1 population for restricted
exophoric reference and ellipsis mainly at observation two. Given the K-1
standard deviations for exophoric reference (.21) and ellipsis (.45) at ob-
servation two, there is a strong likelihood that a single subject in each
case may have accounted for these two, unexpectedly high, K-1 means. Assuff&
ing such, significant differences for the observation factor main effect may
be interpreted largely as magnitude differences.
-53-
me ir um an all me or as us me an or me me me ire um ow osi
R EXO REF
-.14-.
.13-
.12-
.11-e
.10 -
.09 -
.0S -
.0; - 02
.06-
.05-
.04- 01
.03- 01
.02-'...:%\'
.01-
02
.00_ 03 .........-............03
-I
K-11-2
GRADE
.1'
.65
.60
.55
.50 01
.45 03
.40 02
.35
.30
.25
.20
.15
.10
.05
.01
REFERENCE
01
02
ELLI PSIS
K-1 1-2
GRADE
K-1 1-2
GRADE
CONJUNCTION LEXICAL COHESION
.10
.65
.60
.55
.50
.45
.40
.35
.30
.25
.20
.15 03
.10 02
.05 01
.014K-1
GRADE
i
01
02
03
1-2
.70'-
.65-
.60-
.55-
.50-
.45-
.40-
.35-
.30-
.25-
.20-
.15-
.10-
.05-4
.01-
.s.
03
03,'"°°.-- 0202 ...--'''''
01
GRADE
Figure 8. Observation as a Function of Grade for Restricted Exophoric Reference, Reference, Ellipsis,
Conjunction, and Lexical Cohesion in Writing at Suburban School (K-1/1-2)
1) ISO
01
Sex by observation follow-up. Both discriminant function coefficients
and univariate tests of significance indiCted that reference proportiOns
accounted largely for the sex by observation interaction in writing as the
suburban school (See Table 41). hearts and standard deviations in writing
by sex and observation are presented in Table 42. Reference means are graphed
in Figure 9. Malus employed significantly more reference ties at observation
one (.76) than females (.37) and fewer reference ties at observations two
(.38) and three (.40) than females at observation two (.47) and three (.43).
Table 41
Standard Discriminant Function Coefficients and Univariate AMOVAson Use of Cohesion Categories in Writing for Sex byObservation Interaction at Suburban School (K-1/1-2)
Standard Discriminant Univariate F TestsCohesion Category Function Coefficients (2, 40) P.4
Restricted Exophoric Reference .699 1.33 .28
Reference 1.452 53 .006
Ellipsis .748 1.04 . .36
Conjunction .676 .29 .75
Lexical Cohesion .560 . .20 .82
.80- 01
.70-
.60-
.50- 03
.40- 02
. 30-
. 20-
.10-. 0 I
02
0301
Males Females
REFERENCE
Figure 9. Observation as a Function of Sex for Referencein Writing at Suburban School (K-l/1-2).
-55-
a in NI- El do- ow No um no do on um no dui um an a as
Table 42
Means and Standard Deviations of Cohesion MANOVA (Transformed Variables)for Suburban School in Dictation and Writing (K-1/1-2) by Sex, Mode, and Observation
Sex .Mode Observation Exo Ref Reference Ellipsis Conjunction Lexical
M% SD M SD M SD M SD M SD
Males - .02 .06 .47 .25 .03 .19 .17 .10 .36 .16
Dictation .01 .03 .42 .13 .02 .03 .19 08 .38 .13'
1 .02 .04 .42 .20 .01 .02 .18 .07 .41 .17
2 .C2 .02 .40 .06 .02 .0Z .21 .09 .38 .11
3 .00 .01 .44 .09 .02 04 .20 .08 .36 .09
Writing .03 .08 .51 .32 .05 .26 .15 .11 .33 .19
1 .06 .13 .76 .40 .01 .02 .10 .12 .19 .19
2 .02 .03 .38 .21 .14 .45 .17 .13 .36 .17
3 .01 .01 .40 .15 .01 .01 .18 .08 .44 .09
Females .02 .09 . .17 .02 .04 .14 .10 .38 .15
Dictation .01 .02 .41 .08 .01 .01 .21 .07 .38 .08
1 .02 .03 .41 .10 .01 .01 .18 .06 .41 .09
2 .01 .02 .37 .08 .02 .01 .25 .07 .36 .08
3 .01 .01 .43 .05 .01 .01 .18 .0' .38 .07
Writing .02 .12 .43 .23 .02 .05 .08 .09 17 .19
1 .01 .03 .37 .29 .00 .02 ,06 .10 .'2.6 .20
2 .06 .21 .47 .21 .03 .04 .09 .09 .39 .17
3 .0C .0: .43 .17 .04 .08 .10 .08 .46 .16
83
Sex factor follow-up. Discriminant analysis and analysis of variance,
again, were employed as follow-up techniques to probe the significant multi-
variate statistic in writing at the suburban school for the sex factor: F
(5, 16) le 2.95, 2.< .05 (See Table 38). As shown in Table 43, reference
and conjunction proportions contributed strongly to the discrimination for
the s'tx factor; however, only conjunction reached the designate.' alptia level- -
the discrepancy explained by the previously discussed interaction for reference
where observation means veiled substantially but overall reference means for
sex did not differ (See Table 42).
Table 43
Standard Discriminant Function Coefficients and Univariate MIOVAson Use of Cohesion Categories in Writing for Sex Zo: Suburban
School (K-1/1-2)
Cohesion Category
DiscriminantFunction
CoefficientsUnlvariate F Tests(1, 20) 2. 4.
Restricted Exophoric Refer%nce .638 .02 .90
Reference 1.008 3.09 .0:
Ellipsis .370 .42 .52
Conjunction .957 7.54 .01
Lexical Cohesion .527 .66 .43
Observation factor follow-up. The final follow-up for the cohesion
MANOVA in writing at the suburban school (K-1/1-2) employed the usual dis-
criminant and univariate analyses (See Table 44). Lexical cohesion made
the strongest contribution to the discriminant function and was the only
univariate variable to achieve significance. Means and standard deviations
for the observation factor in writing are displayed in Table 39. Lexical
cohesion proplrtions in writing increased substantially across observations
(.23, .38, .45). Table 39 also indAcates conjunction proportions increased
while reference proportions decreased--but not significantly.
Table 44
Standar, .jiscriminant Function Coefficir-PQ and Univariate ANOVAson Use of Cohesiou Categories in Writing for Observation at
Suburban School (,-1/1-2)
Cohesion Categovy
DiscriminantFunction
Co:fficientsUnivariate P Tests(2, 40) .11.<
Restricted Exophoric Reference -0.642 .79 .46
Reference -0.526 2.31 .11
Ellipsis -0.659 1.22 .31
Conjunction -0.503 2.94 .06
Lexical Cohesion -1.220 17.75 .001
DISCUSSION OF COHESION REcULTS
The Transition to Writing: Mode and Grade Comparisons
The most obvious mode differences f. cohesion were magnitude differences
while deploiment differences within categories were minimal. That is, higher
proportions of lexical. cohesi , -onjunction, reference and ellipsis - ,e
employed in dictation than in writing but lower propor ions of restricted
exophoric reference were used in dictation. In dictation, endop'oric ties
were far more prevalent than in writing, despite the fact that the discourse
content was oral. While there were sex and dialect differences in writing,
there were non, in dictation. The major differences reflected in the dicta-
tion data were developmental differences -- differences across observations.
PropJrcion- of lexical cohesion, conjunction, reference, and restricted
exophoric reference varied between modes mainly as a function of development
in writing. Shifts between dictation and writing occurred across observe-
tions--patterns of use of cohesive elements witrin writing gradually coming
to resemble their deployment within dictation. By tne end of grade two,
dictating a story and writing a story embodied nearly identical cohesive
resources.
Development in writing appeared to recapitulate development in dictation.
Comparisons between the two longitudinal populationskindergarten children
observed during their first 16 months of schooling and first-grade children
observed over the same interval--indicated that, while overall proportions
of all cohesive ties were somewhat lower for kindergarten children, differences
between dh and their first-grade counterparts narrowed substantially over
the 16 months. By the time kindergarten childien reached the end of first
grade, cohesion proportions in their dictated and written texts strongly
resembled proportions in the initial dictated and written texts produced by the
old population. Again, the differences between these two longitudinal popula-
tionA gradually declined over observations, but never completely so, suggesting
a developmental lead for the iirst-grade children. As noted for dictation
and writing, sex and dialect differences within both the kindergarten and
first-grade populations were found only in writing. Developmental differences
ordinarily, but not always, favored girls. A history of differential experience
with traditional stories is the most likely explanation for initial distInc-
tions between vernacular and nonveraacular subjects.
Restricted Exophoric Reference. Restricted exophoric reference decreased
in both dictation and writing de both schools, for each grade, in both dialect
86
groups, whether male or female. The decline in exophoric reference across
observations was steeper in writing than in dictation, steeper for boys
than for girls, and steeper for vernacular subjects than for nonvernacular
subjects. This finding clearly suggests that children had generally re-
aligned their intentions and textual functions to the requirements of written
and dictated texts. Given the fact that children were asked to dictate
stories, a task very much like writing, it is not surprising to have found
only minimal differences between modes. Exophoric reference proportions
declined to roughly equal levels for modes-among first-grade pupils but
to somewhat higher levels for kindergarten boys and vernacular pupils in
writing.
Reference. Reference proportions differed between schools, dialects,
sexes, modes and across observations--but not between grades. However, much
of this variability occurred in writing as a function of sex or dialect.
There were significant grade-one increases of reference proportions across
observations in both dictation a..d writing at the urban school. At the
suburban school, reference proportions declined sharply at the beginning of
second grade, but remained at the same level as dictation proportions through-41.
out grade two and approximately equivalent to reference proportions at the
urban school.. Very large, suburban-school differences in first grade, however
can be attributed entirely to the larger proportion of reference ties used
by boys in their written texts--and, quite possibly by just one or two boys,
for the variance associated with the male writing mean was exceptionally high.
Also, urban-school, vernacular females in grade one used more reference ties in
dictation than other groups did, but used about the sole amount of reference
as other groups in writing. Otherwise, reference ties increased moderately in
both dictation and writing across observations. But, what is more important
about reference proportions is that for all observations they averaged from 7.5
-60-87
IA
to 50 percent of all cohesive ties employed in children's texts. If exophoric
reference ties are included, the percentages are much higher at observation
one, particularly in writing. Children at school age appear to have many if
not most relations of text identity well within their competence.
Some tentative conclusions about reference can be drawn. By the end of
first grade, personal and possessive pronouns were rarely used exophorically.
Most remaining exophoria was confined to demonstratives. The issue in the
case of demonstratives appears to have been the problem of distinguishing
between proximals (these) and nonproximals (those) in the anaphoric sense ".,,
of pointing to some aspect of the discourse or to an assumed content of situa-
tion. Comparatives, employed much less frequently than personals and , ..,<
demonstratives, still were employed anaphorically. More confident conclusions
must await detailed analyses of reference ties, and, of course, other cohesion
categories. Frequencies and proportions of specific ties within categories
rather than among cohesion categories will provide more definitive and subtle
indices of development.
Ellipsis. Ellipsis was used sparingly by children in both dictation and
writing. Ellipsis proportionb averaged about two percent of total cohesive
ties in both modes. Substitution, a near relative to ellipsis, was used so
infrequently that it was not included in any of our analyses. Ellipsis, like
substitution, achieves ahesion through relations in words or groups of words.
But, unlike substitution, ellipsis is a relation where something is left unsaid,
yet is understood because of some presupposition in the structure of what must
be infirred, "An elliptical item is one which, as it were, leaves specific
structural slots to be filled from elsewhere." (Halliday & Hasan, 1976, p. 143).
Ordinarily, the presupposed structure is present in the preceding text, and the II
cohesive tie is anaphoric. However, an elliptical structure may be exophoric.
61- 83
So little ellipsis was employed by children in their texts that what
follows is quite speculative.
The variance associated with ellipsis was unusually large, ordinarily
at least twice the central tendency for any given cell. These large variances
imply substantial individual differences. In most, but not all, circum-
stances ellipsis was used in conjunction with dialogue. Also, either one
sex--girls, (grade-one, suburban - school) or the other, boys, (kindergarten,
suburban-school)--in just a single observation accounted for strong increases
in use of ellipsis. Two of several possible explanations seem most likely.
The first is the possibility that some children were experimenting with
dialogue in response to a classroom activity. Commonly, in spoken English,
presupposition is signaled phonologically through tonlc prominence. Tonic
prominence indicates new or contrastive information. The occurance of an
elliptical nominal is nut unusual in this situation. The use of dialogue
in some texts may have given rise to just this opportunity. The secoad
explanation hinges on individual developmental differences. As just noted,
ellipsis and dialogue are naturally linked. Dialogue provides opporLunities
for both nominal and verbal elliptical contrastiveness. Both the oiatogue
convention and its potential for contrastivecess simply may indicate that a
few developmentally advanced children were able to incorporate dialogue in
their texts, and, when they did, proportions of ellipsis increased corres-
pondingly. Either of these explanations as well as others car be substantiated
only through further extensive analysis of cohesive ties within each major
category.
Conjunction. Major differences in conjunction proportions were
associated with modes. Use of conjunction increased across obseriations in
writing but remained relatively stable in dictation. This pattern held for
-62-
89
both kindergarten and first-grade children, although upper-grade proportions
were higher on average. These observation differences for writing were
discussed extensively in Volume 1 (King & Rentel, et. al., 1981). Data from
the kindergarten population and dictated texts provided no grounds for alter-
ing our earlier conclusions. Children's maturing control over conjunction
as a text-forming strategy took the following course. Initial use of con-
junctions in writing was limited to just a small complement of conjunctions:
and, but, so, and then. With these four conjunctions, children were extraor-
versatile in linking clauses and sustaining their written texts.dinarily
Later both imprecision and repetition were replaced by precise conjoined
ties which employed because, soon, although, or, now, plus, and still. In
writing, coordination, subordination, causality, antithesis, sequence, time,
and condition had been organized into a working system of conjoining options,
similar in magnitude and breadth to their dictated texts.
Lexical cohesion. A similar picture to the one just drawn for conjunc-
tion emerged for lexical cohesion. Use of lexical cohesion increased across
observations in writing for all groups in both schools, but in dictation,
lexical cohesion proportions declined moderately for all groups except urban-
kindergarten children. For these children, lexical cohesion increased in both
dictation and writing. Girls used significantly higher proportions of lexical
cohesion than boys. Similarly nonvernacular kindergarten children employed
higher proportions of lexical cohesion in their dictated and written texts
than nonvernacular children at the same level. By observations three, the
end of first grade, vernacular and nonvernacular children employed nearly
the sam2 proportions of lexicai cohesion in their texts, but, by this time,
all children employed more *.exical cohesion ia writing than in dictation.
This finding held for fit grade children as well: by the end of second
grade they employed more :exical cohesion in writing than in dictation.
-63-9
The most likely explanation for both observation and mode differences
is that children had learned to lexicalize meanings to the extent requied
by either dictation or writing--written texts, of course needing full
lexicalization while dictation permitted some meanings to be coded through
prosodic and gestural means. Initial mode differences, where lexical
cohesion proportions were significantly higher for dictacion, suggest that
children were more advanced developmentally ±n the ora: mode, and
uninhibited by spelling, handwriting, and spacing demands, they could focus
their attention on the communicative task at hand. In both dictation and
writing, but particularly in writing, children emphasized the cohesive
relation of co-referentiality through the use of reiteration. Where lexical
items had dual roles to play, an identity role and a similarity function,
children solved the problem of dual function by overmarking the c' esive
relation. Their later texts reflected much greater awareness of textual
relevance through increased use of synonymy, hyponymy, meronymy, and antynomy.
Lexical cohesion proportions across observations revealed several
important dimensions of the transition to writing. Children's greater
awareness of textual relevance is .1 telling indication of the importance they
attributed to semantic foregrounding as a dominant carrier of meaning in
written texts. Children's attempts to overmark the cohesive function under-
scored their sensitivity to the explicit necessities of written texts. The
range and versatility of their lexical ties demonstrated their ability to
express the content of their experiences logically, thus indicating a
greater focus on language's ideational function -- particularly in written
texts that contained higher proportions of lexical cohesion than dictate°
texts. These mode differences ilso suggest that children had realigned
attention and intention to produce texts unsupported by action or context
and to produce texts more sensitive to the needs of a distant audience.
Finally, these increases in lexical cohesion indicate children's growing
recognition that written texts must be internally consistent, that texts
must specify classes of information, as well as relations between, ,within,
and among classes.
Other aspects of lexical cohesion were examined in Volume 1 (King &
Rentel, et. al., 1481). We will not repeat them here. Instead, in a later
chapter, we will attempt to explain the nature of the transition to writing
incorporating both cohesion and story structure in these explanations.
sA,
-65-
1
1
92
Chapter 3
The Transition to Writing: Story Structure for Grade 1-2
Like cohesion, stories have a significant role to play in the transition
from speech to writing. Children frequently tell stories, both old and new,
as they create their first written messages. These stories constitute a
familiar rhetorical structure within which children organize the flow of
discourse. Both Moffett (1968) and Britton (1970) have argued that the
first tentative step children take toward writing is reflected in their ability
to take over a conversation and maintain a topic independent of the prompting
and feedback ordinarily found in dialogue. Britton argues that young children
achieve their communicative intentions through speech but that writing at this
stage in development serves another end: its purpose is to create a tnngible
artifact, a drawing, or a display. Langer's (1953) notion of presentational
symbolism as distinguished from representational symbolism would best
characterize these aims. Children frequently tell a story as they produce
these displays (Britton, 1970). This form of solo discourse between thought
and action embodies both elements of dialogue which are less collaborative
and elements of narrative which are maintained by distinct actions. The cites
children utilize as they develop a text are found not in what an interlocutor
said but in the previous text and in the ongoing constructive actions of pro-:
ducing an artifact. As Vygotsky (1962) noted, language without an inter-
locutor must be consciously directed and sustained to replace the dynamic
guiding quality afforded by a conversational partner. What children learning
to write must grasp is how to take what is implicitly obvious in the context
93
-66-
and render it explicit in text. Cook-Gumperz (1977) characterized this trait
as the ability to appreciate language as a structure separate from action.
At school age, children have learned the underlying structure of stories
(Johnson6 Handler, 1977; Glenn & Stein, 1978). These structures appear
to be nearly fully represented in memory, for, when asked to recall stories
which, have been randomly organized, children produce a stereotypic or canon-
ically organized version'of the tale (Johnson & Handler, 1977; Glenn & Stein,
1978). Fur*her, there is some evidence that four and five year-old children's
t,
descriptions of common event sequences such as eating lunch at McDonald's11
(Nelson, 1978)' rely heavily on schematic organization. This suggests a gradual
acquisition of a story schema beginning with script -like chronicles which
continue to grow in structural complexity up to age ten and beyond (Botvin &
110 IISutton-Smith, 1977), culminating in well-formed, episodically organized
structures--girls earlier than boys (Sutton-Smith et al., 1975; Duggan, 1977).
A
If, indeed? memory for events and instances is so-organized, then story
schemata may constitute one of the fundamental cognitive bases for the
rhetorical scaffolds employed by beginning writers.
Both Winograd (1979) and Halliday (1973) maintain that such discourse
schemata do provide guide or models for integrating language into texts.
One such patters is narrative. Halliday (1973) arzes that children develop
conceptions of what language is and haw it works--that is, "relevant models"
which represent a pattern of discczrse. We think that such conventionalized
models figure heavily in the design of,childrea's beginning narratives and
expect that fairy tales and folk tales provide a rhetorical framework for11
beginning writers. But, the extent to which such schemata guide production
is not really known--however appealing or likely such a notion might be.
94-67-
1
There is, of course, evidence that fairy tale and folk tale elements are
represented in the original stories children tell and write. Rubin and Gardner
S.
(1977) argue from their data that children acquire a general frame for fiction
starting at about three years of age which they theri differentiate into
specific story genres. .4y four years of age, children appear to have partially
represented the "frame" fhr fairy tales ( Rubin & ahrdner,1977). By six,
stock characters such as witches ana fairies appear in their written and
dictated stories (Applebee, 1978). Oral narratives produced by children
demonstrate that action elements very much akin to Propp's functions--plot/*
units - -do Fleed characterize the organization and structure of cldren's
fantasy narratives (Botvin, 1977; Botvin & Sutton-Smith, 1977).
Botvin and Sutton-Smith (1977) reported that many, but by no means the
majority of their subjects, told fantasy narratives resembling the fairy tales
analyzed by Propp (1968). Using a modification of Propp's morphological
functions, Botvin and Sutton-Smith observei that the complexity of compo.ent
action sequences in children's narratives increased in a direct relationship
with age. Starting with nuclear dyads, that is, two logically related actions
children progressively expanded and elaborated these basic structures into
fully-embedded zomplex plots. It is not clear, however, what role if any
familiar folk and fairy tales played in providing these children with relevant
models of fantasy texts and to what extent such models guided their early
productions. Most narrative plots analyzed by Botvin and Sutton-Smith involved
either a lack, and its liquidation or a villainy, and its nullification. These
elements are identical to those posited by Propp--lack and lack liquidated;
and villainy coupled with villainy nullified. In Propp's morphology two
additional pairings, struggle coupled with victory, and difficult, task paired
with its solution, augmented the obligatory functions of lack and villainy.
95-68-
This coincidence between children's narratives and the formal attributes of
fairy tales as set forth by Propp suggests that, at some point in learning ..o
compose, many if not all children employ a narrative schema quite similar to
tales they have heard and read.
Fairy tales have a highly conventionalized plot structure (Propp, 1968).
Favat (1977), who compared various popular tales ranging from Perrault to the
Grimms and Anderson, observed that these tales have an extraordinaKily pre-
dictable structure and bear a striking similarity to their Russian counterparts
analyzed by Propp. Even Propp, speculating that fairy tales may have a common
origin, made the following observation:
Yet one still feels inclined to pose this question: if allfairy tales are similar in form, does this not mean that theyall originate from a single source? The morphologist does nothave the right to answer this question-. At this point he handsover his conclusions to a historian or should himself become ahistorian. Our answer, although in the form of a supposition,is that this appears to be so. However, the question of sourcesshould not be posed merely in a narrowly geographic sense...The single source may also be a psychological one. Much hasbeen done by Wundt in this sphere. But here also one must bevery cautious. If the limitation of the tale were to be ex-plained by the limited faculties of human imagination ingeneral, we would have no tales other than those of our givencategory, but we possess thousands of other tales not resemblini,fairy tales. Finally, this single source may come from everydaylife. (p. 106)
Other literary structuralists (Todorov, 1971; Maranda & Maranda, 101; Bremond,
1970; Dundes, 1964; Levi-Strauss, 1963) have explored the constitutive princi-
ples which define the narrative form. Despite differences among them, their
analyses bear certain fundamental similarities. They all identify a principle
of order or "succession," and both Propp (1968) and Todorov (1971) set forth
a principle of transformation--though Propp did not incorporate the principle
of transformation formally into his analysis of structure. These structuralists
also identify functions, elements or units ,which are indispensable or essential
-69-
96
1.
1
to the narrative. These elements generally include a beginning marked by some
initial state cf satisfaction or equilibrium, a complication or degredation
of this initial state, a recognition of this change in state by the protogo-
nist, an action which repairs or remedies the complication, and a restoration
of equilibrium. None, however, completely fits the tales written for the
enjoyment of children nor the tales written by children. Propp's analysis,
however. does have-the advantage of breadth and delicacy.
Hasan (1980) has argued that the stories children= compose are a separate
genre of fictional narrative which can be described in terms of five obligatory
elements: (1) placement, (2) initial event, (3) sequent event, (4) final event,
and (5) finale. She also proposed other nonobligatory elements found fre-
quently in children's tales such as rituals, attributions, habitu-1 actions,
and relations to characters. King, Rental and Cook (190) compared Hasan's
analysis of structure with Propp's and found that they correlated rather well
(.65) when only obligatory elements were included in the analysis of narrative
texts produced by six-year olds. Leondar (1977), who also analyzed children's
narrative texts, concluded that children at the age of five or six produce
texts that include an Initial state of affairs, an event that disrupts this
state, a counteraction to reverse the disruption, and a restoration to the
original state. Thus, the stories that children tell and write bear an
appreciable resemblance--at least in terms of structure--to the various ways
in which stmcture in fairy tales and folk tales has been described, Leander
puts it this way:
The constructive powers of the author and the re :onstructiveones of the reader may be assumed to spring. from a common source.On both counts, then, the development of narrative competence inearly childhood invites examination. (p. 173)
Fairy tales, of course, comprise only one of the many genres of storles
children encounter both in and out of school. Why should fairy tales and folk
-74-
97
tales be singled out as rhetorical models for beginning writers? First, the
literature, as indicated above, provides evidence that young children's texts
mirror many of the elements typically defined as elements of fairy and folk
tales. Second, when children retell stories, even stories in which the
underlying grammar or structure has been violated, their retellings are
biased toward a prototypic or canouical form (Johnson & Mandler, 1977; Glenn &
Stein, 1979). Third, fairy tales and folk tales delight and engage children
unfailingly (Favat, 1977), and as Favat speculated, they probably do so
because of their highly. conventionalized structure. On these grounds, it
is reasonable to assume that fairy tales in all probability are rather
well-represented in memory by school age. To the extent that they are, we
expect that such tales are fundamental rhetorical guides for beginning writers.
But how do such guides function in the production of a text? Our notion
is that abstract story elements provide a range of options for selecting and
organizing events in a temporal sequence revealing and emphasizing relations
between and among characters and events (Leondar, 1977). They also provide
a reservoir of states, complications, and repairs of an abstract sort--frames--
to be,propositionalized as events and roles (characters) to express the ideas
contained in these frames (King & Rentel, 1979) . The young storymaker must
sustain a narrative in some cumulative way. Regardless of variety, the story
.eller must produce a schema containing both necessary and sufficient elements
of a story. Such frames would provide the basis for cumulating units either
additively or in parallel. Even with a minimum of rudimentary elements,
through repetition, a narrative could be sustained indefinitely (Botvin &
Sutton-Smith, 1977). Finally, particular sets of elements can be combined in
parallel or in tandem affording the storymaker opportunities for thematic varia-,
tion. The storymaker can give dimension to a story in the making.
These various perspectives converge on the notion that conventionalized
models of text figure in the design of children's narratives. But why study
children's narratives? Why not study the entire range of discourse children
are capable of producing at school entry? First, even though rather sparse,
there is at least a growing literature on the production of narrative texts
by children at school age. The importance of having an existing literature
from which hypotheses and methodology could be derived is self-explanatory.
Economy alone would have been sufficient justification for focusing upon
a single genre some of whose attributes and dimensions had already been
characterized. Then, of course, with genre controlled, an'important source
of variation could be examined without need for further complicating an already
complex set of logical and statistical comparisons. In addition, this existing
literature gave rise to our expectation that children would be more likely to
produce texts of greater length and nuance in narrative form than in other
discourse genres. Finally, of necessity, research in school contexts must
conform to the ongoing life of a classroom. Our problem at the outset was to
select variables and manipulations which fit nicely into this context, yet
constituted reliable and theoretically significant aspects of writing develop-
ment to study. The structure of fictional narratives seemed to rest at the
intersect of these points.
Analysis of Story Structure Data
Population sampling, story structure definition and analyses, genre
classification, scoring and interscorer reliabilities were identical to pro-
cedures described in Volume 1. For the reader unfamiliar with Volume 1,
detailed descriptions of these procedures are given in Appendix A'. Only the
designs comparing story structure variables differed in the present studies,
not in kind but in nuance and detail. These differences are described in the
next section.
93-/2 -
Both multivariate and univariate analyses of variance were employed for
story structure comparisons. For the multivariate analyses, as with cohesion,
computer program CANOVA (Clyde Computing Services, 1973) was used. This pro-
gram tests for significant differences with Wilk's likelihood ration trans-
formed to Rao's approximate F. Significant multivariate differences were followed
up with univariate analyses of variance.
Number of function, function types, and moves served as dependent
variables in six complementary multivariate analyses of variance performed
on the story structure data. In the first of these analyses, 144 scores for
each dependent variable were organized into a mixed design where sex (6 males
and 6 females) and dialect (6 vernacular and 6 nonvernacular) served as
between-subjects comparisons, and where discourse contexts (retelling and
dictation) and observation periods (Spring, 1979; Fall, 1979; Spring, 1980)
constituted the within-subjects comparisons. This study was designed to cow-
pare factors within the urban school setting. Similar design arrangements were
employed in a second analysis whose purpose was to compare the urban with the
suburban school controlling for dialect. While only middle class children
from the two schools were compared, the two populations did differ on the
Index of Status Characteristics with F (1, 20) = 5.39, 21... <.05. Children
from the suburban school averaged from middle to upper-middle class on the
"index" (M = 32.25; S.D. = 4.41) while those from the urban school averaged
somewhat higher scores on the scale (M = 38.33; S.D. = 7.47). The two pop-
ulations had been equated on the scale at the outset of the study, but because
of subject mortality and replacement, this initial equality was lost necessftat-
ing a school comparison. For this comparison, dependent variables were
organized into a 2X2X2X3 fixed design where sex and school were the between-
subjects factors and where modes and observations were the within- subjects factors.
11
-73-
A third multivariate analysis of variance, then, was employed to examine only
the suburban school. As before, number of functions, functiontypes and moves
were organized into a fixed design with one between-subjects comparison, sex
(6 males and 6 females), and two within-subjects comparisons (contexts and
observations).
Three additional multivariate analyses of variance focused upon dictation.
Retelling was removed as a comparison in order to obtain a clearer view of
dictation over the three observation periods -- retelling differences having
potentially spurious origins in the variance associated with a priori story
differences. In all other respects, design goals and arrangements were
identical to those reported above. These studies were reported in Volume 1.
Likewise, an additional multivariate analysis of variance focused upon
texts produced by a sample of subj4cts who were able to compose unequivocal
fictional narratives. Just 14 subjects were able to do so at mid-grade one.
This number rose to 27 by the end of grade two. The point of this multivariate
analysis was to obtain developmental data controlled rigorously for genre.
OLher kinds of texts were excluded from this analysis to eliminate genre as
a source of contaminating variance. Consequently, other comparisons..such as
sex, socioeconomic class, and dialect slipped from our grasp because, as might
be expected, not all groups were equally represented in this new sample. Thus,
dictation, retelling, and writing were compared in a fixed design over.the
three observation intervals--a comparison where modes of discourse and observa-
tions comprised within-subjects factors.
Multivariate analyses of variance were used in these studies to reduce
the chance of drawing erroneous conclusions. Repeated univariate analyses
over the same data would have entailed a potential "alpha error." Testing
the significance of differences obtained in repeated analyses over the same
-74-101
subjects using dependent variables which were conceptually and mathematically
related had to make allowances for the fact that the scores obtained for each
variable were undoubtedly correlated with each other. A test of significance
was selected that would take these correlations into account thereby avoiding
false conclusions about the probability of having observed a true difference- -
an "alpha error." Multivariate analysis of variance (MANOVA) makes such
allowances, and requires of the researcher an a priori decision regarding
true differences among all the variables of interest before tests of sig-
nificance may be conducted for these same variables individually.
Then, too, the concept of structure we had advanced had more than a
single dimension to it Story structure as defined by 'functions, function
types and moves was three-dimensional. To account for how social class,
school, sex and observations contributed to this composite, a method of 0:01W'
parison had to be selected which would be sensitive to the direction and
magnitude of these relationships assuming; of course, conconmiitant variation
among the three variables that defined story structute. Again, multivariate Janalysis of vn-iance permitted such comparisons.
We had decided that when significant multivariate (MANOVA) story structure
effects were obtained for a factor or factors, these effects would be followed
up with univariate analyses of variance (ANOVA) rather than any of a number
of possible multivariate techniques that might have been employed.' The
decision to employ univariate analysis of variance as a follow-up procedure
was based upon the notion that for each of the dependent variables which made 11
I Because a clear a priori construct could be posited for cohesion, we employeddiscriminant function analysis as a follow-up technique to explain significantdifferences obtained in the multivariate analyses of variance of cohesionvariables. Unlike story structure, theoretical relationships among the five v .
categories of cohesive ties were amenable to rigorous linguistic interpreta-tion and relatively clear theoretical explanation.
102-75-
up the construct story structure, we had developed hypotheses which could be
tested with reasonable prospects for explaining observed differences. That
is, for the measure, number of functions, arguments had been advanced and
assumptions had been examined c.itically resulting in a determination that
number of functions was a viable and reliable index of sustaining power, and
that, likewise, number of function types indexed storymaking range, and that
moves measured complexity. The relationship of each of these dependent
variables to the various independent variables we were probing could be given
a tenable hypothetical explanation. But when functions, function types and
moves were combined conceptually as the construct story structure, we had
neither appropriate data to support the efficacy of such a construct, nor
riviorbus theoretical backing to permit disciplined explanation of fundings
that might result from our analyses. On these grounds, a multivariate follow-
up procedure 'was ruled out for the story structure studies. In short, we
simply concluded that the problem of interpreting relations between independent
and dependent variables in the case of story structure was larger than the
current state of our data and our knowledge could handle within a multivariate
perspective. Thus, we opted to follow up significant effects from multivariate
analyses of variance, a procedure that protected against "alpha error," with
univariate analysis of variance, a procedure that conservatively limited the
scope of our conclusions and generalizations.
Story Structure: Mode Comparisons
As noted above just briefly, one set of analyses compared texts produced
by a sample of 14 subjects (Grade 1-2) who were able to write fictional
narratives when requested to tell or write a story. These analyses, of course,
were designed to obtain developme-tal comparisons controlled rigorously for
-"genre. To illustrate the methods employed in our longitudinal study, these genre -
cont rolled comparisons will be presented below.
103-7b-
Texts that children produced, that is, retellings, dictations, and
written stories, were analyzed and scored for number of functions, number of
function types and number of moves by pairs of independent scorers. Different
pairs of scorers were employed for each of the three production contexts.
Diterscorer reliabilities were established for retelling: .89; for dictation:
.93; and for writing: .85. Over a sample of twenty texts, two scorers
achieved a 78 percent agreement level on individual functions and two others .
achieved an 89 percent level of agreement. Differences among scorers were
resolved and function definitions were ref/tied. 'Having satisfied ourselves
that our scoring procedures were essentially reliable and that we could
reach satisfactory levels of agreement on scoring individual functions, we
opted for economy's sake to have texts scored by a single scorer and checked
by another for possible re-scoring. While double scoring of texts would have
led to some increased accuracy, scoring language protocols is enormously time-
consuming and therefore expensive, as anyone who has done it knows. Gains
in accuracy must be weighed against costs, for obviously beyond a certain
point small gains in accuracy become extremely costly. Reliabilities of the
magnitude already attained suggested we had reached that point; therefore,
we substituted a procedure wherein every text was scored and checked and
differences resolved where necessary.
For those texts which were uniformly fictional stories,,four multivariate
analyses'of variance were employed to compare story structure variables. Each
multivariate analysis of variance was followed-up by pertinent univariate
analyses of variance. In the first of these multivariate analyses of variance,
a two-factor, repeated-measures design, functions, function types and moves
served as dependent variables while discourse context and observations served
as the two within-subjects comparisons comprising the fixed factors in this
-77-
104
design. The three additional multivariate analyses of variance were employed
to probe each discourse context separately, the reasons for these added analyses
to be given below. Each of these, in turn, was followed up by univariate
wereof variance where significant multivariate effects were obtained.
Results of the. two-factor repeated-measures MANOVA are given below in
Table 45. Discourse contexts .end observations were .the within-subjects
factors in this design, an for both factors as well as for the first-order
interaction, significant multivariate effects were obtained. Means and
standard deviations are given in Table 46.
Table 45
Two-Factor Repeated Measure Story StructureMANOVA by Discourse Contexts and Observations
Source df dfHYP dfERR
Discourse Context (A) 2 6.00 48.00 24.18 .001
2.00 24.50 13.69 .001
SA 26
Observation (B) 2 6.00 48.00 8.06 .001
2.00 24.50 1.92 .168
SB 26
Discourse Context 4 12.00 132.58 3.17 .001
6.00 123.34 .73 .623
2.00 102.00 .63 .534
SAB 52
TOTAL 112
105-78-
Table 46
Maans and Standard Deviations of Functions,Function Types and Moves by Discourse Contexts
and Observations
Discourie Observation Functions Function Types MovesContext M SD M SD 14 SD
Retelling 16.59 5.09 12.38 3.65 2.07 0.75
1 15.64 7.07 11.21 3.81 2.14 0.862 17.36 2.41 11.2:, 0.83 L.93 0.273 , 16.19 4.93 14.64 4.31 2.14 0.94
Dictation :8.86 6.99 6.07 3.49 2.02 1:35
8.21. 8.56. 5.36 3.69 1.86 1.41
2 9.79 7.74 6.14 2.83 2.14 1.75
3 8.57 4.42 6.71 3.97 2.07 0.83
Writing 3.36 2.66 3.12 2.41 0.93 0.51
1 1.79 2.23 1,57 1.87 0.50 0.522 3.71 2.89 3.64 '2.82 1.07 0.273 4.57 2.14 4.14 1.70 1.21 0.43
1.
1
Follow-Up ANOVAs For the Discourse Contexts Factor
To determine the nature of the significant multivariate differences
obtained for the discourse context factor, univariate analyses of variance
(ANOVAs) were performed on each of the dependent xariables. Table 46 displays
1the means and standard deviations for these dependdnt variables by discourse
contexts and obsetvations and Table 47 presents the ANOVA summaries. Sig-
nificant differences were obtained among discourse contexts for each of the
three dependent variables.
As shown in Table 46, and demonstrated by Tukey post hoc comparisons of
means, children incorporated significantly more functions and function types
in their retellings than they included in their dictations and written stories
106-79-
lar 111111 VW MI am 1111114 la' UV ON III MB MI SW MS ea ail' WU 'Ile MI
Table 47
Follow-up Univariate ANOVAs on Functions,Function Types, and Moves by Discourse Contexts
1
01
SOURCE .if MS F p4 df MS F p< df MS F P4
Discourse Context (A)Error (SA)
2
26
1857.5927.66
Functions
.001 2
26
Function Types
.001 2
2617.560.71
Moves
.00167.16* 940.14 80.79*11.64
24.75*
TOTAL 28 28 28
* Geisser- Greenhouse conservative F test using reduced degrees of freedom (1, 13) is significant at p< .01.
1071 oc-;
4
and significantly more functions in their dictations than in their written
texts. Children also included significantly more moves in their retellings
and dictations than they did in their written texts. As expected, retell-
ings were more easily sustained than dictations and dictations more easily
sustained than written stories. Likewise, retellings included a greater
range of functions than did dictations and dictations a greater range of
functions than writing. In both the retelling and the dictation contexts,
Children achieved roughly the same level of complexity--at least as complexity
was indexed by moves, but in their written stories children produced
significantly less-complex stories than they produced in either of the other
discourse contexts.
Follow-Ups ANOVAs for the Observation Factor
In addition to the significant multivariate test statistic for the
discourse contexts factor, the MANNA on the story structure variables
produced a significant Wilks' lambda criterion for observations: F (6.00,
4k.00) a 8.06, 4.001. After removal of effects associated with the leading
root for observations, no significant discrimination remained. To determine
the nature of the observation differences relative to the story structure
variables, follow-up ANOVAs were performed on the three story structure depen-
dent variables. Means and standard deviations for these variables are in
Table 48. Only the follow-up ANOVA on function types produced a significant
Table 48
Means and Standard Deviations of Functions,Function Types, and Moves by Observations
Observations Functions Function Types Moves
K SD M SD M SD
1 8.55 8.58 6.05 5.11 1.50 1.222 10.29 7.44 7.02 3.95 1.71 1.11
3 9.98 6.47 8.50 5.68 1.81 0.86
-81- 103
univariate effect for observations. This effect is displayed in Table 49.
Tukey post hoc tests demonstrated that observation-three texts, overall,
contained significantly more function types than texts produced at observation
one.
Table 49
Follow-up UnivAriate ANOVA onFunction Types by Observations
Source df MS F p. <
Observation (B) 2 64.02 9.34* .001
Error (SB) 26 6.85
TOTAL 28
* Geisser-Greenhouse conservative F test using reduced degrees (1, 13) is
significant at R.< .01.
Follow-Up ANOVAs for the Discourse Context by Observation Interaction
Table 45 indicates a significant multivariate effect for the first-order
discourse context by observation, interaction. Again after removing effects
associated with the leading root for this first-order interaction, no
significant discrimation was observed for the remaining two roots. However,
follow-up ANOVAs performed on the three story structure variables failed to
turn up significant univariate interaction effects for any of the three
dependent variables. Coupled with the significant effect for the discourse
contexts factor, this finding of a significant multivariate interaction led
to a decision to perform separate multivariate analyses of variance for each
discourse context by observations as a way of trying to tease out the meaning
of the interaction. Since clearly each of the three contexts had been demon-
strated to be significantly different from the other, multivariate analyses
of variance were in order to probe eabh context as an individual entity.
-821-1
The Retelling Context
Scores. from the three story structure variables were organized into a
one-factor repeated-measures multivariate analysis of variance (MANOVA)
wherein observations served as the within-subjects factor. Table 50 shows
that a significant multivariate effect was obtained for observations. When
effects associated with the leading root for observations were removed,
no significant discrimination remained. This significant multivariate effect
for observations was then followed up by univariateiralyses of variance
(ANOVA) performed on each of the three dependent variables for story
structure: functions, function types and moves. These analyses (ANOVAs)
produced a significant effect for the observation factor only on the function
types dependent variable (See Table 32). Post hoc comparison of means
(Tukey's HSD Procedure), which are shown in Table 51, demonstrated that a
significantly greater range of function types were included in retellings
at observation three than were included in retellings produced at observations
one or two. The number of function types included in retellings produced at
observations one and WO, however, did not differ significantly.
Table 50
Story Structure MANOVA in RetellingContext by Observations
Source df dfHYP dfERR
Observation (A) 2 6.00 48.00 7.88 .001
2.00 24.50 2.44 .108
Error (SA) 26
TOTAL 28
111
Table 51
Means and Standard Deviations for FunctionTypes' in Retelling Context by Observations
Observation Function TypesM SD
1 11.21 3.812 11.28 0.833 14.64 4.31
The Writing Context
Results from the MANOVA on the writing data are shown in Table 53.
This MANOVA summary table indicates a significant multivariate test statistic
Table 52
Follow-up Univariate ANOVA in RetellingContext on Function Types by Observations
Source df MS <
Observation (A) 2 53.74 9.08* .001
Error (SA) 26 5.91
TOTAL 28
* Geisser-Greenhouse conservative F test using reduced degrees of freedom(1. 13), is significant at R.< .01.
for the observation factor. Removal of effects associated with the leading
root for observations demonstrated that no significant discrimination remained.
This significant multivariate test statistic for observations was followed up
by univariate analyses of variance (ANOVAs) on the three story structure
dependent variables. Means and standard deviations for these three dependent
variables by observation are given in Table 54.
Table 53
Story Structure MANOVA inWriting Context by Observations
Source df dfHYP dfERR Q. <
Observations (A) .2 6.00 48.00 5.25 .001
2.00 24.50 1.41 .263
Error (SA) 26
TOTAL 28
Table 54
Means and Standard Deviations of Functions,Function Types, and Moves in Writing. Context by Observations
Observation Functions Function Types MovesSD M SD SD
1 1.79 2.23 1.57 1.87 0.50 0.522 3.71 2.89 3.64 2.82 1.07 0.273 4.57 2.14 4.14 1.70 1.21 0.43
As shown in Table 55, significant univariate effects were obtained for
the observation factor in writing on all three dependent variables.
1. Follow -Up ANOVA on Functions: As shown in Table 55, a significant univariate
effect was obtained on the functions dependent variable for observations.
Tukey's USD Procedure indicated that the mein for observation three was
significantly larger than the mean for observation one. A significantly
larger number of functions was incorporated in children's written stories at
the end of second grade than ware found in their written texts at mid-grade one.
113
NE ilk tse Oa in 1111 IP WHO MI 1.1 Oa 0111 MR (111111
OW,
Table 55
Follow-up Univariate ANOVAs on Functions, Function Types and Movesin Writing Context by Observation
. SourceFUNCTIONS
df MS 7- E.<eUNCTIONTYPESof MS F E.<
MOVESdf MS F E.<
Observation (A)
Error (SA)
TOTAL
2 28.50
26 4.14
28
6.88* .004 2 26.02 7.81*
26 3.33
28
.002 2 2.00 13.00** ,001
26 0.15
28
* Geisser- Greenhouse conservative F test2, < .05.
** Geisser- Greenhouse conservative F testE. < .01.
114
using reduced
using reduced
degrees of freedom (1, 13), is significant at
degrees of freedom (1, 13), is significant at
111
2. Follow-Up ANOVA on Function Types: Table 55 indicates that a significant
effect for observations was obtained on the dependent variable function types.
Tukey post hoc comparison yf means (Table 54) showed that observation three
differed significantly from observation one and that observation two also
differed significantly from observation one. Children incorporated a
significantly wider range of function types in their written stories both
at the beginning and at the end of second grade than they did in the stories
they wrote at mid-grade one.
3. Follow-Up ANOVA on Moves: Finally, Table 55 indicates that a significant
text statistic was obtained for the observation factor on the dependent
variable moves. Again, Tukey's HSD Procedure was employed to compare means
(Table 54). These comparisons showed that only observation three and
observation one differed significantly. .Stories written at cle end of grade
two were significantly more complex than the stories children wrote at mid-,
grade one.
The Dictation Context
As indicated by Table 56, no significant multivariate effects were
obtained in the dictation context for the observation factor. Consequently,
no univariate analyses of variance (ANOVAs) were employed to follow up this
MANOVA.
Table 56
Story Structure MANOVA inDictation Context by Observations
Source df dfHYP dfERR
Observations (A)
Error (SA)
TOTAL
2
26
28
6.002.00
48.0024.50
1.200.23
.321
.795
116-87-
This combination of multivariate analyses of variance taken together
with univariate follow-up procedures constituted the principal techniques we
employed to obtain and probe developmental comparisons. One additional
explanation of procedure may be in order at this point. When we employed
'univariate analyses of variance having repeated measures, Geisser-Greenhouse
conservative F tests using reduced degrees of freedom were incorporated
into these analyses. to repeated-measures designs, an assumption of homo-
geneity of covariance is made. To the extent that population covariances
are heterogeneous, the F test is likely to be biased positively. The con-
sequence is that the likelihood of a Type 1 error will increase as a function
of the number of repeated measurements. That is, a probability exists that
when the null hypothesis is true, the test of significance will reject it as
false. The Geisser-Greenhouse adjustment protects a designated alpha rate
even in the face of drastic violations of the homogeniety of covariance
assumption. This protection for one-factor repeated-measures designs, for
example, is accomplished by reducing degrees of freedom to 1 and n-1 (Geisser
and Greenhouse, 1958). Thus, in all cases where we employed repeated-measures
designs, the Geisser-Greenhouse correction was applied.
Are Stories Relevant Production Models for Beginning Writers?
Earlier we argued that action elements similar to those identified
by the Russian structuralist, Vladamir Propp, were often 6und in children's
fictional narratives (Botvin & Sutton-Smith, 1977, Rubin & Gardner, 1977,
Applebee, 1978). Based upon this evidence as well as evidence that what
children recall of stories skews toward a prototypic story form (Handler &
Johnson, 1977; Stein & Glenn, 1978), we hypothesized that these action elements
provide relevant discourse models for beginning writers. We argued that fairy
tale and folk tale models influence the production of stories in three ways.
First, the young storymaker must sustain a text in some cumulative way. The
storyteller must produce a text containing the necessary elements of a genre.
Just a relative handtul of such elements would provide children with a flexible,
cumulative basis for creating story, plots, a genre most children know well.
The key to sustaining a text is repetition. ArLy function or group of
functions may oe repeated until sufticienc volume has been created. For
Children, sustaining a text without toe interactive support of a conversational
partner is itself a cnaliensing task. oimultaneousiy, the struggle to imagine
and create elements oi a plot, characters, texture, and a consistent narrative
stance coupled with the mysteries of spelling, handwriting, and spatial
organization also make enormous demands on processing capabilities. Repeti
tion of functions potentially can restrict the range of action elements.to
be managed, and thus simplify the problem of text formation. Plot, character
relations, narrative stance and texture may be held in check while continuing
to maintain a story line. More attention can be devoted to actually producing
a written artifact without sacrificing important narrative intentions. The
following excerpt from a beginning second grader illustrates this use of
repetion.
(1) Once there was a wicked old witch
(2) her name was treetop
A (3) now_she was a very very wicked old witch
(4) she lived on witch main street
(5) one day when she was walking down she saw a little girl
(6) and she said in heT mind that she was going to eat her
(7) so the next day she.took a little walk and saw the girl
(8) the girl said do you think she is going to eat me
(13) so the next day when she took her walk her mom came with her...
Once setting information has been established in lines (1) through (4), the
repetition of the functions departure and villainy (notential) sustain this
text throughout the remaining units. Repetition of any given function or
combination of functions, thus, provides a means for sustaining a text.
Having this capability would provide children with an important asset in
helping them make the transition from face -to -face oral text production to
solo written production.
Second, the availability of underlying abstract functions provides a
limited range of options from which an almost. limitless variety of stories
. Jr
can be generated. Particular sets of options serve as defining features for
a genre such as the fairy tale. This ability to specify genres constitutes
an important advantage of functions over more abstract story grammars. In
addition, functions, linked as they are to the actions of characters, define
not only the relations between a protagonist and events that make up a plot,
but have the potential for defining relationsamong characters and thus the
capacity for specifying multiple relations with events.
Third, logical pairings such as lack and lack liquidated or villainy
and villainy nullified, comprise nuclear combinations or "dyads" which Botvin
and Sutton-Smith, (1977) have described as basic building blocks in the
development of narrative competence. These pairings, when combined in tandem
or in parallel or when embedded one within the other, afford the storymaker
opportunities for thematic reflection, permutation and variation. The
principal advantage to be gained through such pairings is that of dimension-
ality. Using pairings such as these, children would possess the ingredients
to create rather complex stories without making excessive demands on memory.
Very substantial differences among discourse contexts for all three
story structure variables, demonstrated that production was clearly affected
by input factors and the availability or, perhaps, accessibility of infor-
mation stored in memory. Retelling texts contained roughly twice as many
functions and function types as dictated texts and about four to five times
as many functions and function types as written texts. Both retellings
and dictations contained significantly more moves than wttttstritories.
Quite obviously and expectedly, producing a written text makes enormous
demands and places severe restrictions on composing capabilities. Over
observations, a period of 16 months, only a significant overall effect for
function types was obtained. When each discourse context was analyzed
separately, however, significant effects for all three dependent variables
were obtained only for tne writing context. In the retelling context. just
function types increased significantly over observations, and in the dicta-
tion context, no significant differences were obtained over observations for
any of the dependent variables.
The finding that number of function types increased significantly over
observatiOns for both the retelling and the writing context suggests that
developmental increases in the ability to comprehend and recall stories are
accompanied by corollary increases in production capabilities. Recalling
that the stories children retold at observation one and at observation three
contained an equal number of function types, the differences observed in
their retellings over this 16-month period cannot be attributed to input
stories. Thus, comprehension and recall abilities increased significantly
over this period and production abilities improved concomitantly.
Correlations between retellings and written texts on number of function
types ranged from low (.34) at observation one to moderate (.53) at observa-
tion three; thus, these two tasks increasingly shared variance associated with
an expanding range of function types. it seems reasonable to conclude that
-41- 12u
1
the same developmental factors which were responsible for differences in
retellings over observations were also responsible for the wider range of
function types found in children's written texts over this same period of
time. the viriable function types, presumably, is rooted in exposure to folk
and fairy tales or their fantasy counterparts found in television cartoons
and serials as well as motion pictures.
Some appreciation of how much children's tales came to bear a clear
resemblance to traditional folk and fairy tales can be gleaned from the
following story written by a child at the end of second grad.
Unit Function
(I) once upon a time there was a bunny named Benjie Setting
(2) and she had magic powers Magical Agent
(3) one day she was walking in the woods Departuie
(4) and a boy bunny appearedIt
(5) and they went together for a walk Translocation
(6) and a man appeared with a big net Reconnaisance
(7) and he got the two bunnies and went in a big ship Villainy
(8) poor bunnies ri
(9) they were caught now
(10) but right then the girl bunny tripped the man Struggle
(11) and they got free once again Villainy Nullified
(12) so the boy bunny thanked the girl bunny for Reward
saving him
(13) the boy bunny asked the girl bunny to marry him Equilibrium(Propp's wedding)
(14) and she said yes
(15) so they had six baby bunnies
(16) and they lived happily ever after
The resemblance between ':his second grader's story and a fairy tale is
a strong one. The range of functions included in this tale exceeds the range
of functions typically produced by late-second graders. Yet, in other respects
such as length, cohesive airs, number of functions, and complexity, this story
is fairly typical of texts written by children near the end of second grade.
The evidence we have obtained about the role of stories in beginning
writing development is far from decisive. It does provide Weak but plausible
grounds for believing that the comprehension and representation of stories in
memory constitute rudiments of a rhetorical schema for composing stories during
the beginning phases of writing development. Over time, children in our sample
increasingly built narrative structures predicated upon function-like action
units. As they wrote longer stories, they included more functions and a greater
range of function types in their tales. The correlation betqeen story length
as measured by number of T-units, and number of functions over observations
ranged from .61 at observation one to .92 at observation two and to .70 at
observation three. Correlations between text length and range of function types,
as might be expected, closely paralleled those for number of functions; observa-
tion one, .54; observation two, .90; and observation three, .75. Both sets of
correlations indicated that functions and function types were measuring the gamek.)
or nearly the same underlying capability relative to text length. But function
types appeared to be a somewhat more sensitive developmental indicator than
functions. These correlations with length suggested that as texts increased in
length, tt.e.: manifested greater breadth and range of storymaking capabilities
and increased sustaining power on the part of the developing writers we studied.
We think that familiarity with stories played a substantial role in this
development.
-122-93-
Chapter 4
The Transition to Writing
Several years ago we began a longitudinal study of 42 kindergarten and
42 first-grade children seeking to discover how their text-forming strategies
changed as they entered school and shifted from producing mainly oral texts
to producing written texts. It is now four years later. We have analyzed
data from the first two years of the study and are continuing to sort out
and analyze data from the last two. More specifically, the text-forming
strategies we wanted to study were those employed by children to relate
various layers of meaning encoded in texts and those employed to build a
rhetorical structure. Most earlier research rm primary and intermediate
grade children's writing, aside from the work of Graves (1975) and Clay
(1975). had focused upon syntax (Hunt, 1965; Loban, 1976; O'Donnell, Griffin &
Norris, 1967) or phonology (Read, 1971). Of course, Hildreth (1936) had
studied the development of rudimentary aspects of children's writing from
three to six years of age as did Wheeler (1971) for kindergarten children's
beginning efforts to write. Our purpose, however, was to describe how chill-
dren use cohesive ties to relate strands of meaning in their spoken and
written texts.
In an earlier work (King & Rentel, 1979) we sketched a theory which
attempted to account for several facets of beginning writing development.
Briefly, we argued that what children bring to writing in the way of oral
language, early concepts of the functions of written messages, their exposure
to and their sense of stories, fundamental differences between written and
spoken language and particular factors which influence production sach as memory,
context, and audience affect the way children execute their communicative
intentions. We noted that the decisions involved in an utterance entail
different plans, depending upon the unit being planned. Basing our argu-
ments on Chafe's work (1977a, 1977b), we hypothesized three levels of
planning--schematizing, propositionalizing, and categorizing--corresponding
to text units (schema, proposition and category).
At the first level of planning, schematizing, the speaker's or writer's
task is to plan chunks large enough to convey a coherent unit of meaning, but
small enough to constitute a unit of memory for particular instances and
events. Story schema, as structured according to "functions" by Propp (1968)
was chosen as the most productive means for exaraing the schemas employed
by children.
Favat (1977) had argued that children's interests in traditional tales
arises from their expectations of the relative invariant structures of such
stories. In other words, children had a basis (schema) on which to make pre-
dictions. Moreover, work by Botvin (1977) and Botvin & Sutton-Smith (1977)
.indicated that children's stories did in fact include functions as defined by
Propp.
We further argued that fundamental distinctions between speech and
writing-which we will explore in detail later--would constitute the bases for
expanding and refining these planning capabilities. Briefly, planning at each
level would require realignment from the implicit, shared, dimensions of speech
in situational contexts to the textual demands of writing where all meanings
must be made explicit in the immediate text. We expected this realignment
would be most clearly revealed through cohesion (Halliday & Hason, 1976),
language's major resource for linking elements of a discourse. Cohesion is a
range of possibilities by which meanings in a text may be related through
124-95-
reference; substitution, ellipsis, conjunction and lexical cohesion. These
devices specify the natrue of information to be retrieved when interpreting
any particular segment of a text--which at the beginning stages of writing
development children may assume to reside in a context of situation rather
than in their text. We expected variations in'obntext, content, audience
and purpose would affect the kind and distribution of cohesive resources
children employed at all three levels of planning. In particular, the propor-
tion of references to context of situation should decrease while those confined
to the text should increase as a function of learning to write. The need to
embody relevant meanings within the written text, coupled with developmental
increases in vocabulary (Clark, 1973, 1979), we expected would lead to expanded
use of lexical cohesion, which, in turn, would affect propositionalizing and
categorizing. We argued, therefore, that an understanding of beginning writing
development necessarily should include an exploration of cohesion in children's
oral and written texts.
Another major element of our theory was premised on young children's
early writing intentions gleaned from their prior exposure to written texts
and their sense of the purpose or function of writing. One kind of written
text with which most children are intimately acquainted is the story. Given
that a great deal of their experience with writing has been through stories,
children are likely to assume that one of the major functions of writing is
to tell stories. We anticipated that children incorporate this assumption into
their understanding of general language functions (Halliday, 1973), that writing
serves an ideational function--language's ability to characterize experience,
and a textual function--language's ability to specify internally consistent
relations within a text, comprehensible without reference to anything outside
the text. Our notion was that, for children, stories are prototypic texts
clearly associated with writing whose function, form, and internal relations
are reasonably well known to them. As a natural extension of this knowledge
base, in their early attempts to write children will favor stories for schema
planning. Of course, there are pragmatic considerations as well. Given that
schooling is expected to produce literacy, language arts instruction usually
offers children the opportunity to tell and write stories in the primary
grades. Both schooling and their knowledge base incline children to plan
schema in terms of familiar story and folk tale structures. As children
learn to make stories of increasing detail and.complexit Y, and acquire a
deeper appreciation of the nature of written texts, their schema planning
can be expected to become more formal, deliberate and conscious. In the
beginning of writing development, though, such planning probably is intuitive
and unconscious.
Developmental Aspects of the Transition to Writing
Learning to write has its roots deep in oral language develdpment.
Children's first tentative steps toward writing summarize this extraordinary
accomplishment and retrace elements of earlier growth, all the while dis-
covering and exploring similarities and differences between talking and writing.
At the same time their experiences with language are extending toward print,
children have begun also to experience a world beyond the home, the world of
neighborhood, school and new acquaintances. And with this expanding world
comes new opportunities to revise and add knowledge of every sort to nourish
further growth. But there are constants, too. The familiar stories of bed-
time crop up in new school guises as do routines, chores, denials, and
assorted frustrations. Together, these ingredients comprise the context for
learning to write.
Of the 36 kindergarten and 36 first-grade children studied, only 13
first-graders and five kindergarteners, 22 percent of all children, were able
to write a story at the outset of the study. Sixteen months later, only one
child still could produce no written text whatsoever. However, nearly all
children were able to retell stories and most were able to dictate stories
right at the outset of the study--somi dictations, of course, based upon
stories children had heard seen in film or on television. Many of the first
texts children attempted to write were little more than statements or pictures
with labels. Other children wrote texts but their texts had many of the ear-
marks of oral interactions with an assumed conversational partner. The first
written texts children produced averaged less than thiee T-units in length.
The texts they produced 16 months later averaged 10\T-units in length. Dic-
tated texts, on the other hand, averaged 19 T-units\in length at the outset
and 35 T-units in length 16 months later. Both numbe:of cohesive ties and
number of function correlated strongly with text length. As stories grew
longer and incorporated a wider range of story functions so also did they
includi increasing proportions of lexical cohesion and conjunction while de=
clining correspondingly in proportions of restricted exophoric reference.
By and large, dictated texts changed in the same way, although earlier than
did written texts. Concommitently, story retellings reflected children's
increasing ability to comprehend and recall stories. The range of story
functions recalled was approximately twice the range children included in
their dictated stories. Thus, the prediction that oral texts comprise a basis,
both textual and structural, for learning to write received clear but modest,,
support from these broad mode differences which diminished across observations.
Story structure data and cohesion data also indicated that children do,
indeed, realign basic oral language strategies in the process of learning
to write. From the story structure findings, it can be concluded th4t children
gradually acquire a knowledge of underlying story functions evidenced increas-
ingly, by their broadening storymaking capabilities in writing. This breadth
\
is first manifested in their retellings, then, dictations, and eventually
in written texts. Their stories not only came to reflect greater sensitivity
to structure but awareness of necessary logical relationships among story
functions as well. These abilities were accompanied by increasingly precise
use of conjunctions to achieve relations of coordination, subordination,
causality, sequence, time, and conditionality--precision essential to 4pecify-
ing relations among story functions. Both sensitivity to structure and con -
joined logical indicators among functions, argue for a conclusion that children
have learned to emphasize the ideational function of language in their texts.
The same close link exists between use of other cohesive devices and
story functionsthat is, between form and texture (Pappas, 1980). Ambiguous
or exophoric referent items sometimes obscured who was participating in an
7action resulting in an uninterpretable story function. The foll 'ing story
partially used as an example in Chapter 3, illustrates this problem,
1 once there was a wicked old witch
2 her name was Treetop
3 now she was a very very wicked old witch
4 she lived on witch main street
5 one day when she was walking down she saw a little girl
6 she said in her mind that she was going to eat her
7 so the next day she took a little walk and saw the girl
8 the girl said do you think she is going to eat me?
9 now she was not sure she was going to eat her
10 but she had a big feeling she was
11 so she went on and on and on
-99-128
In lines 9, 10, and 11, the ultimate referent of "she" is unclear. And
the potential villainy is equally unclear. The story continues as follows.
12 when she got home she said to her mom "mom today when I took awalk I saw a witch--a real true witch"
13 so the next day when she took a walk her mom came with her so hermom would see her too
14 and they did see her
15 her mom got scared out of her wits
16 and they tore home and started to moan
17 one day the little girl got sick
18 and she threw up
19 that was the only time she was nice because she made her better
20 and she gave her a kiss '
21 but she shouldn't have done that because all of a sudden sheturned into a witch
22 and her mom did everything she wanted her to
"She" in lines 19, 20, and 21 probably refers to the witch who kissed the
little girl causing the little girl to get well. On the other hand, "she"
might also refer to the little girl and "her" to the witch whose villainous
nature was bettered by the little girl's kiss. Either interpretation is ride
more difficult by the previous ambiguous referents in line 9, 10, and 11 in
which the villainy was obscured. Too, note that conjunctions, which might
have specified appropriate logical relationships between functions, are not
sufficiently precise to do so.
Contrast the "witch" story with the story excerpt which follows.
1 once there was a bear who lived in the woods
2 he was lonely
3 he had only a few friends
4 one day he went to racoon's house
5 his friends were there
6 his friends were a squirrel, a racoon and a deer
7 he had no other ones
Personals, demonstratives, and comparatives combine to specify identity
unmistakably. Setting information and the story function, departure, are
equally-clear. These two examples demonstrate that structure and texture
are inseparable. Development appears to be premised upon pooling information
from multiple sources.
The second example also illustrates the writer's awareness of the need
to establish textual relevance in writing by expressing nearly all meanings
through words. The cohesive relations of hyponymy and reiteration--embodied
in friends, squirrel, racoon, and deer--coupled with explicit identity rela-
tions, realized through personals, possessives and comparatives, combine to
make this story segment clear and coherent. This text, and others like it,
are typical of the orchestration of lexical cohesion and reference in late,
grade-two, written texts. Each set specifies a consistent relationship to
a character, thing, event, or act and the classes and related classes of
information to be found in the text. Proportional increases in endophoric
reference and lexical cohesion indicate a fundamental awareness of the explicit
textual demands of writing.
As children acquired a broader range of story functions in memory, their
written stories increased in detail and complexity. Their deeper appreciation
of structure also coincided with their expanding range of cohesive options
giving support to the notion that writing development is integrally linked
to children's ability to organize and interpret knowledge through dyr .c,
-101-1 3 u
prototypic schema large enough to convey coherent units of meaning, but
sufficiently discrete to constitute memory units for events or stereotypic
actions. These schemata eventually are differentiated into fixed and vari-
able elements defining a genre. Our hypothesis was that these conventional
models of discourse undergird planning in the sense of selecting and
organizing information to be expressed and understood in a text. From this
larger chunk, pertinent details are assigned a role and theme in events
and actions constituting the larger schema. Cohesive devices, in turn,
specify the information to be retrieved when interpreting any given segment
of text, thus linking schema and propositions. Cohesive ties, therefore,
also function as variables influencing selection of words and phrases. At
best, our data suggest these planning dimensions. Indeed, production was
affected by input factors and the availability, or perhaps accessibility,
of information stored in memory. As demonstrated in the examples above,
factoring out role and theme were obviously linked to cohesive resources
available'to the writer. And, textual identity and relevance required
selection of lexical items from rather narrowly defined semantic fields.
The pattern of developmental increases'we observed weakly supported our
explanation of planning; however, only experimental evidence can settle
these process questions. Our descriptive data merely point to an interesting
set of possibilities.
Factors Contributing to Developmental Variation in the Transition to Writing
The factors which influence learning to write are highly interactive.
These interactions range from contextual to constitutional contingencies as
learning cumulates over time. Our goal was to separate important sources of
variance from this flux. Our expectations for establishing unambiguous causal
131-102-
relations were modest. Instead, we hoped to describe the emerging histories
of children whose backgrounds and schools differed substantially, through
successive, regular well-timed observations which incorporated slight
manipulations in discourse modes and tasks. These manipulations were de-
signed to capture broad, global, characteristic similarities and differences
among the populations studies. We expect that our subsequent analyies of
cohesion and text structure will tease out basic trends in writing develop-
tk.
went. The studies reported here deal only with global characteristics.
These global characteristics were: school, grade, sex, socio-economic status,
and dialect. Discourse mode and grade-level variation have already been
discussed. In this section we will consider school, sex, and dialect
variation:the latter, tied securely to socio-economic status.
School variation. Both kindergarten and first-grade populations at the
urban school differed from their counterparts at the suburban school with
only middle class children included in the comparison. These two levels of
school populations differed only on the writing comparison, suburban school
males in kindergarterfaing higher proportions of reference than suburban
females or urban males and females; suburban school first-graders using
higher proportions of lexical cohesion than urban school first-graders.
Differences in kindergarten reference proportions can be attributed confi-
dently to atypical scores for suburban males--an effect that is easily
explained. All suburban, kindergarten males produced a written text at ob-
servation one while only two suburban females did, only one urban male
produced a written text and all urban females did. Suburban boys included
mainly "setting" information about characters -- information which required
specification of text identity entailing use of reference ties. Urban girls,
on the other hand, also included mostly setting information in their texts,
-103-
-132
a
but, unlike suburban boys, urban girls also employed reiteration to achieve
textual relevance which entails use of lexical cohesion. Suburban kinder-
garten boys, therefore, employed very high proportions of reference ties in
their first written texts while urban girls employed roughly equal proportions
of reference ties and lexical ties in their first written texts.A
Overall, first-grade and kindergarten children at the suburban school
employed higher proportions of conjunction and lexical cohesion in their
written texts than urban children. To the extent that school differences
are implicated in these effects, the following factors may be involved; At
the suburban school, every effort was made to integrate reading and writing
around focal, interests and long units of study. Some small group and indivi-
dual instruction was given in reading. A great emphasis was placed on
literature and the use of a range of books, both fiction and nonfiction, in
all learning activities. Literature was also studied for itself. Teachers
frequently read aloud to children, discussed books with them and often
organized books for study around a common theme, concept, author or illustrator.
This curriculum provided substantial opportunities for purposeful exposure
to written texts, particularly stories. On the other hand, individual teachers
at the urban school had considerable lattitude in determining the type of
literacy instruction employed. Emphasis was on skills--in word recognition,
handwriting, and spelling. A wide range of textbooks, audio-tapes and
duplicated materials was used in teaching reading--usually at the discretion
of the teacher. Toward the end of the school year, more attention was given
to the content of children's writing, to exposing children to clusters of
books and stories, and to reading aloud to children and telling stories.
Lexical cohesion and conjunction differences may simply be a reflection of
_104_133
the suburban children's earlier, more consistent, purposeful exposure to
stories which resulted in a developmentally earlier realignment of textual
functions and strategies.
Sex variation. Girls and boys differed only in writing and mainly at
the first grade level. These first-grade differences mainly involved use of
lexical cohesion. At the kindergarten level, boys at the suburban school
used higher proportions of reference in their written texts, a finding which
was discussed in the previous section on.school differences. First-grade
girls employed lexical cohesion proportionately more than first-grade boys,
whether because of rate differentials in vocabulary acquisition, or because
of specific differences between boys and girls in text-forming strategies
employed to achieve textual identity and relevance. Answers to these. ques-
tions must await further detailed analyses of lexical cohesion.
Dialect/socio-economic status variation. Jithin the urban school,
middle-class, nonvernacular dialect children were compared with lower-class
vernacular dialect children. Cohesion differences between these populations
without exception, were limited to written texts. Within the kindergarten
population, middle-class, nonvernacular children employed higher proportions
of l'xical cohesion in their texts than lower-class, vernacular children.
When both kindergarten and first-grade populations were included in'the
analyses, overall means for writing were higher for nonvernacular children on
both ellipsis and conjunction proportions. When only first-grade children
were included in the analysis, reference and ellipsis means in writing were
higher for nonvernacular children than vernacular children.
Mean differences for lexical cohesion in the urban kindergarten popula-
tion stem from the fact that fewer nonvernacular children wrote texts at the
outset. Only one vernacular, kindergarten child wrote a text at observation
one while four nonverr. .lular children wrote texts. Mean dialect differences
for lexical cohesion in the urban, kindergarten, written texts persisted over
observations. Whether these differences stem from acquisition differentials
or from differences in wa>.; of creating textual relevance cannot be determined
from comparisons of lexical cohesion proportions. To explain these differ-
ences, frequencies of particular lexical ties, each type having specific
implidations, will be compared in later studies. What car be said now is that
these two populations of kindergarten children appear to employ different
text-forning strategies--nonvernacular kindergarten children relying more
on lexical devices than vernacular children to produce written texts. The
finding of no lexical cohesion differences between thee:' kindergarten pop-4
ulations in dictation probably rules out acquisition differentials as a ten-
able explanation. The more likely explanation is that these vernacular
and nonvernacular children employed different lexical means for creating
textual relevance. The absence of leetal cohesion differences in writing
between first-grade dialect groups indicates, however, that, whatever the
sourcfis, lexical cohesion differences subside by the end of grade two.
But, dialect group differences for reference and ellipsis did emerge
in first-grade writing. The nonvernacular group mean for reference
proportions in writing was .46 while the vernacular group mean for reference
was .24. Again, no dialect group differences were olerved in dictation.
Ellipsis means were also higher for the nonvernacular group in writing, but
no dialect group differences for ellipsis, were observed for dictation.
Ellipsis proportions were very low and, as noted earlier, were largely con-
finee to texts containing dialogue. In all likelihood, this style difference
135-106-
accounts for the dialect group difference for ellipsis. Reference differences
are not as easily explained. Differences were large and are probably an
indication of different strategies for achieving text identity, but only
further analyses of particular reference ties can reveal just what these
differences are and what their ultimate explanation is.
Dialect group differences for story structure are difficult to
interpret. More recent analyses (See Appendix D) of kindergarten story
structure data--the dependent variables, functions, function types and moves--
indicated only dialect-group differences.for retellings in the kindergarten
population for number of functions and number of function types. Consistent
first-grade, dialect-group differences were indicated for all three dependent
variables in both dictation and retelling. Patterns of difference over
observations indicate that both vernacular and nonvernacular groups experienced
comparable increases but maintained initial relative differences between them
for all three dependent variables. Retelling differences, particularly, may
have been either a result of middle-class, nonvernacular children having had
broader exposure to stories in general, or the result of riddle- class,
nonvernacular children having ' pi broader exposure to the kinds of folk and
fairy tales read to children participating in the study.
Longitudinal Studies of the Transition to Writing
The structure in which scientific knowledge is created is relatively
straightforward and simple. The scientist merely asks, "How are two things
related?" "Under what conditions do they affect one another?" "What is a
reasonable explanation of these relationships and effects?" The scientist's
logic, attitudes and traditions of peer review are his major defenses against
error. Aside from the ways in which conditions and things are controlled,
scientific methods differ mainly in nuance. At the heart of the enterprise
is reasoning of the form, "if p, then q." At the end is an empirical test
of the validity of the scientist's explanations. The most creative aspect
of science is determining what p and q are.
Longitudinal research is one method for asking questions and submitting
them to empirical tests. Its advantage and disadvantage is that things come
to the researcher in their natural states where both things and their
relationships are difficult to sort out. Since a necessary first step in
science is to determine what to observe, that is, to delineate the dependent
variables of the phenomenon under study, some consideration must be given to
how these variables are to be selected. Our arguments above, of course, led
to the selection of cohesive ties and elemental functions in fairy tales as
variables salient to the transition from spoken to written texts. But aside
from the arguments given, were there other reasons that influenced our
selection? Both cohesion and story structure promised to be sensitive to the
sorts of manipulations permissible in the on-going life of classrooms--sensi-
tive because we could reliably observe them in the protocols children produced
and we had a fairly full grasp of their theoretical significance. The
conditions for learning to write were very clearly defined by the existing
curriculum in the schools we studied. Obviously we had little choice in that
matter. But we could establish or easily devise ways of assessing a child's
status with respect to these variables and thus characterize the emergence of
the ability to write under specified conditions. These conditions we
attempted to describe fully and clearly.
The factors which influence learning to write are enormously complicated
and highly interactive. Unravelling the sheer number and kinds of interactions
ranging from contextual to constitutional contingencies as they cumulate
over time is a staggering problem. Our initial aim was to separate relevant
independent variables from this flux. We had little or no hope of determin-
ing causes. What we did have, however, were children whose histories and
achievements typically differ dramatically--poor inner city black children
and affluent suburban white children--and in that inner city magnet school,
a population of affluent, middle class, white children. By focusing upon
their emerging histories through successive, well-timed observations which
incorporated ever-so-slight manipulations, we hoped to ensnare rather large
global attributes at first. Subsequent studies, designed to sift the data
from these observations through finer and finer analyses, will provide us
with a fuller more complete picture of the factors that contribute to growth
in writing ability. As stated earlier, we have identified the following
global characteristics school, grade, sex, socio-economic status, dialect,
and discourse context.
The ultimate purpose of any kind of educational research, of course,
is the compilation of knowledge necessary to guide educational theory and
practice. The goals of educational research, regardless of type, are to
minimize harmful or wasteful experience and maximize the probability of optimal
development for each student. When all goes well, there even may be the
potential in educational research for making contributions to basic knowledge.
Research on writing shares these purposes and goals. Like educational research
in general, research on writing shares also many of the fundamental diffi-
culties encountered in school related inquiry.
Not the least of these problems is that of obtaining reliable and rep-
resentative samples of data. Most learning contexts and all school settings
are complex interactive social entities. We would like to propose two
safeguards against error arising from these difficulties. First, that to
the extent possible, observations or measures of attributes of writing be
set within the fullest theoretical backing available. Where theoretical
backing is weak or lacking, then the most stringent protections available
should be applied to the analysis of data and great care taken to restrict
generalization. Second, that multiple measures or observations within
varying contexts--that is, sampling of several sorts from the population
of situations where writing is likely to occur--be taken. Our work thus
far has demonstrated to us that there is an ebb and flow to writing
development linked intimately to situations and population groupings. And
there'are individual eddies in this swirling current. There is an old and
familiar nursery verse that begins, "Monday's child is..." and ends with
"Sunday's Child..." We have learned to expect this variability in each child.
The major drawback in our research, as in all longitudinal research, is
the inability to draw strong causal inferences from our data. It is not
possible in any sense with our data to do more than describe the staLus of
our various populations at given points in a developmental trajectory. The
value of longitudinal data is the control exercised over within-subject
differences. Stable differences among subjects can be identified in a way
that is reasonably free of this within-subjects error. The power of our data
and of longitudina] research in general is its firm link with reality and
its potentially strong developmental effects. These links and effects can
blaze a trail for more compelling and causally persuasive experimental research.
In this sense longitudinal studies of writing development constitute a pro-
ductive approach to theory and inquiry: longitudinal research can and often
does serve as an important source of reasonable experimental hypotheses. And,
133-110-
at a very practical level, longitudinal studies set out clear developmental
beacons that teachers may follow in shaping their expectations for
curriculum and instruction.
Chapter 5
Transition to Writing: A Case Study
This report is a study of the development of Jwan's writing abilities
from tie end of kindergarten through the end of first grade. Jwan is a
black, lower class child. When data were first collected, Jwan was five
years old and a member of a kindergarten-first grade classroom. This
group of students was housed in an open area and contained approximately
seventy-five students and three female teachers. During the 1979-1980
school year Jwan moved to another kindergarten-first grade classroom which
was in an open area that housed approximately fifty students and two
female teachers.
The data for this case study were collected in the Spring of 1979, in
the Autumn of 1979, and in the Spring of 1980. One story retelling, one
dictated story, and one assigped writing were collected during formal data
collection in each period. Samples of unassigned or continuous writing
were also collected during each period.
The development pf text cohesion and story functions will be described
for each mode. The development of these devices vat also be compared
across modes and will be related to the child's attitudes about writing
and to the context in which the writing occurred.
The Development of Text Cohesion
Story Retelling,. Juan's first story retelling was short, only six
T-units in length', and was characterized by a high incidence of exophoric
reference, or those references to persons or things which were not
specifically stated in the text. This use of exophora assumes the audi-
ence's knowledge of information not contained in the text. Jwan was told
that the listener to his retelling had no prior knowledge of the story,
but he was unable to make all references explicit within the text.
Jwan did use two endophoric references, or references to items within
the text and two repetitions of lexical items. He connected his ideas with
the conjunctions and and then. Infrequent use of cohesive devices and the
frequent use of exophoric references produced a brittle text with little
cohesion. The following example will illustrate this point.
1 the wolf gonna eat the duck
2 and he said there was a piece of cheese
3 and there was a moon
4 and the duck got a spanking
5 and she asked him to be a good little duck
6 then he went to bed
Figure 10. Retelling of Squawk to the Moon Little Goose
142
1
Jwants second retelling was longer than the first and slightly more
coherent. His use of exophoric and endophoric references were equal, and
he used more lexical repetitions than the first sample. He continued to
connect his ideas with simple additive conjunctions. This retelling is
more understandable than the first.
1 they had no food
2 they only had a piece of bread
3 and then the little girl went out in the forest
4 and a old lady gave her a magic pot
5 ((sp: unknown)) "stop boiling little pot"
6 and then the pot kept on boiling
7 then she ran down the street
8 and then the girl said the word
9 and then it stopped
10 and then they had a lots of food for everyone
Figure II. Retelling of The Magic Porridge Pot
The final retelling indicates a reversion to the earlier means of
developing a text. Exophoric references outnumber any other single cohesive
device used, and they approach the total number of all other cohesive
devices used. There are only three endophoric references. Cohesion is
established through lexical repetition and simple additive conjunctions.
-114-
143
1 the boat sunk away
2 and the boy was eating the salt
3 and all the children was looking at the giant
4 and the little boy was fibbing about him
5 the giant say you can tell all you want about me
6 ana he little boy got married
7 and the people tore up the boat
8 and the boy was staring at them
9 then the boy went sailing off with the boat again
Figure 12. Retelling of Salt
This pattern of text cohesion indicates a lack of awareness of
audience and an inability to create an independent text. 10 fully
understand many of the references that Jwan used in his retellings,
the audience must have knowledge of the original story.
Jwan consistently depended on conjunctions (and, then) to link
units of his text. The number of lexical items that were repeated in
his texts indicated that he had added reiteration to his text-forming
strategies.
Jwan's story retellings are brief and are lacking the necessary
information to make them coherent. Over a year's span Jwan showed little
development in his ability to create cohesive retellings of given stories.
Story Dictation. Jwan's first and last story dictations were similar
in length and in content. They were also similar in that the cnly exophoric
references used refer to Jwan himself. There is a greater use of endophoric
reference in his dictations. since both of these dictations are brief, there
are only two conjunctions in each piece. These continue to be simple addi-
tive conjunctions. Jwan also used lexical repetition resulting in brief but
cohesive texts.
A
1 I play with my daddy and my dog 1 me and my friend Artez rideon big wheel .,nd bike with one
2 and my dog chase me and bite me hand
3 I run out on the road" 2 and we always race
4 I like to play with my friends 3 and I always beat the race
5 and one is Bobo
Figure 13. Dictation: Spring-Kindergarten; Spring-Grade One
The story Jwan dictated in the autumn of 1979 was much more elaborate
and more fictional than the examples just given. Within this story, the
number of endophoric and,exophoric references was equal. However, many of
the endophoric references were ambiguous so the meaning became obscure in
parts of the story. In the following example Jwan used simple additive
conjunctions and lexical repetitioas to create textual cohesion.
145
1 they was fighting
2 and Joker got knocked out
3 then they went back to their bat cave
4 then they got in their batmobile
5 then they zoomed of to the Joker's hide out
6 then Batgirl jumped in and started fighting withBatman and Robin
7 and then Batman and Robin Superman Spiderman theywere all having cars Superman Spiderman Batman andRobin
8 and they zoomed over a house eeerrrkkk
9 then they were home
10 then they slid down the post and turned back into* Batman and Robin and lefty the Joker home again
11 ' then Joker in the middle
12 then Catwoman jumped in
13 and she tried to jump across that thing
14 and she fell down in it
Figure 14. Dictation: Autumn-Grade One
In Jwan's dictated stories, endophoric and exophoric references were
used equally as often. He relied more on lexical repetition and conjunc-
tions as text-forming strategies. It is interesting to note the increased
length of text and'the imaginative,style of Jwan's Autumn dictation. This
story is a sharp contrast to the Spring, 1979 and the Spring, 1980 dictations.
Assigned Writing. It was impossible to analyze Jwan's assigned writing
samples for cohesion because he produced no texts. His writing consisted of
drawings and experimentation with letter forms. The Autumn, 1979 samp'e
contained the most writing. It was a list of words copied from parts of his
class room.
-11,7-
146
b
TFO ice 4-VA15 .05INA \ co51 rA
IC05
°Cr01,11-2111
Figure 15. Writing: Spring-Kindergarten
0
Although no text was produced, Jwaa's writing showed that he had a
good concept of letter and word, and that he was aware of print around
him. As the following examples show, Jwan's early writing was more of
an effort to make a product than a means of communicating a message.
This type of writing represented one of his first steps in becoming a
writer.
.can's last writing sample consisted of words copied from books
combined with some of his own words. He sequenced words; showed his
knowledge of left-right directional' y, spacing, and top to bottom
sequencing. He was also beginning to create original spellings for
words and there was some effort to convey a message. Jwan seemed to
be at the very beginning stages of writing development. He WAS slowly
moving toward developing written texts.
f -hagr-E6tic60.91 sP ON
\O-F-ebtq
frif car ° "-"-grictiNtce
t
Figure 17. Writing: Spring-Grade One
Summary Swan's story retellings and dictated stories were marked by
his use of exophoric reference. This type of reference seemed especially
prevalent in his story retellings. The use of exophoric reference indicated
that iwar. was not clearly ewer,: f the needs of his audience and the textual
function of writt i language. He seemed unable to decenter enough to provide
his audience with the information that he possessed. Much of the cohesion
in Jwan's stories resulted from his use of conjunctions and lexical repeti-
tion. Jwan's ability to retell and dictate stories surpassed his ability
to write his own stories. However, his writing demonstrated progress
toward creating written texts. Jwan's oral abilities exceeded his written
abilities at this stage of development.
The Development of Story
Story Retelling Jwan's retellings were characterized by use of rela-
tively few story functions. The actions of story characters seemed
fragmented and unrelated. In Jwan's first retelling of Squawk to the Moon
Little Goose, certain aspects of story function appeared. There was a
statement of attempted villainy which was not clearly resolved in the
retelling. Punishment occurred but it was punishment of the story hero,
not of the villain. This rebuke of the hero led to restoration of
et.uilibrium at the end of the retelling (See Figure10).
The second retelling incorporated a more complete sequence of story
functions. Jwan clearly established the function of lack in the first two
wits of the retelling. Lack was followed by departure of the hero and an
encounter with a donor through which the hero received a magical agent.
After the receipt of the agent, the sequence of functions broke down. The
final line of the retelling liquidated the original lack and restored
equilibrium. Figure 11 demonstrates the sequence of functions just described.
Jwan seemed confused as to what events occur after the hero receives
the magic pot. Functions Lname unclear until final equilibrium was
restored. Jwan was able to resolve the initial problem and to end the tale.
The final retelling that was collected from Jwan had no clearly defined
story functions. This production was a Last of isolated fragments of the
story. There was no clear beginning or ending, and there was little-relation-
ship among the events listed (Sec Figure 12).
Jwan's story retellings did show an awareness of story functions. He
seemed interested in ending his stories in a state of equilibrium even if
the body of the story contained unrelated functions and no clear sequence.
Jwan was beginning to create story functions, but at this stage, he often
abandoned the functions before they were rebolved.
Story Dictation Only one of Jwar's dictated pieces was a story which
contained story functions. This story was collected in Autumn, 1979 (See
Figure 14). There was not a well defined beginning to the story; it opened
with a struggle without any preceding lack or villainy. The struggle moved
to a departure of the characters, then, to another struggle followed by
the charZ:ters moving to another setting. The story ended with one of the
characters experiencing a misfortune. There was no precipitating lack or
villainy necessary to form a complete tale. This dictation was a series of
beginnings of functions which lacked the necessary structure and sequence
that define a tale.
Assigned Writing No story functions appeared in Jwan's writing.
Summary Jwan was beginning to include story functions in his story
retellings and dictations. More functions and more complete sequences of
functions occurred in Jwan's retellings of stories, but they were altered
and reorganized. Only one of Jwan's dictations contained crude functions,
but no functions appeared in his writing. As with cohesion, Jwan's oral
abilities to use story functions outstripped his written abilities at this
point in his development.
- 122-151.
Unassigned Writin4.
Jwan did not write very much on his ow.. His early writing eoncisted
of drawings with labels and lists of copied words and letters. we
occasionally copied sentences from a chart that the teacher prepared daily.
He also copied words and phrases that were posted in the classroom. By
Jut'', 1980, Jwan was beginning to create original sentences. These often
consisted of repetitive writing like the following example:
I like pupsI like aI like fish
This type of writing was beginning co appear on the same page as copied
words and phrases. Jwan was still in the very early stages of writing
development, but he was beginning to break through to literacy by copying
writing which existed in his environment and by adding his own words and
letters to that writing.
Attitudes About Writing
When interviewed, Nan revealed that he did not like to write and that
he did not think he could write. When asked to give reasons for writing,
he could give none. He stated that he did not write very often and could
name no occa5lon to write in school. Jwan considered writing in school to
be a purposeless, meaningless activity that he should avoid.
Jwan did state that he wrote at home with his father, and they would
sometimes write notes to his teacher. When asked why he wrote to his teacher,
Jwan stated matter of factly that he needed to tell his teacher something,
but she was not there. This statement clearly showed Swan's awareness that
writing can convey meaning; however, oral language was preferred over written
-123: 152
language. Juan had difficulty relating writing that he did at home to
the kind of writing that he did in school. They were almost two different
systems to his way of thinking.
These attitudes may aid in understanding Jwan's development in story
retelling, story dictation, and writing. These tasks may not have been
meaningful to Jwan; therefore, Jwan may have not related them to his
private, purposeful concept of writing at home but may have related them
to his concept of writing at school.
Context for Writing
For the 1979-1980 school year, Jwan was a member of a kindergarten -
first grade group of fifty children and two teachers. This was a highly
verbal classroom. Directions and interactions were usually verbal. In
this classroom, written language was secondary to oral language.
The writing that children were exposed to was primarily captions or
simple statements. Pictures with labels hung on the walls. Posters and
banners with a few words or phrases decorated bulletin boards. Printed
rules were posted. There was very little writing that was elaborate or
expressive. Ueither the children nor the teachers needed writing to work
in the classroom. Therefore, little writing developed in the classroom.
The classroom teachers verbally valued children writing in the classroom,
but they found it difficult to operationalize their values because their
priorities were focused on helping the children complete predetermined tasks
and to work responsibly in learning centers. Often a child's request for
a teacher to read what he had written was ignored or denied. The teachers
did not realize the importance of responding specifically to the child's
-124-
153
composition in terms of meaning. Many children would not initiate their
own writing because they felt other behaviors were more pleasing to their
teachers. Teachers occasionally encouraged children to write, but they
often did not encourage the children's efforts.
This was a busy, productive classroom where children were free to
talk and move and interact with each other. Books were present in the
classroom, and a story was read daily, but books and writing did not
pervade the curriculum. Printed materials were not related to the ongoing
activities in the classroom. Books, printed materials, and children's
own writing were rarely used as resources for the children. If used, these
materials usually remained in the teacher's domain,a
Little writing occurred in this classroom because writing was not
integral to the life of the classroom. Because of the many demands of
the curriculum on these teachers, they were unable to create an environment
where writing was necessary and purposeful. There were few reasons to
write in the classroom so little writing occurred.
Conclusion C.;
How a child develops his writing abilities seems to be closely related
to his mastery of oral language, his ability to create a sustained oral
narrative, and his knowledge of story structure. The child's own attitudes
about himself and writing, and his opportunities to interact with written
language and to write al3o seem to influence how writing develops.
Jwan's abilities to retell and to dictate stories exceeded his abilities
to write his own stories. His retellings and dictated stories improved from
Spring, 1979 to Autumn, 1979 and then seemed to regress to their original
level in Spring, 1980. His writing developed very slowly, but did progress
steadily from Spring, 1979 to Spring, 1980.
Jwan's reasons for writing and his opportunities to write were limited
within the classroom context. Limited opportunity and lack of confidence
in himself as a writer influenced the quality and quanitity of Jwan's
writing. Schooling must provide both reasons and rewards for writing.
Time to write without interruption, and a place for writing to cceur are
important to writing development. Jwan's developing writing abilities
must be supported by a nurturing context if he is ever to develop into a
mature writer.
BIBLIOGRAPHY
Allerton, D. J. The sentence as a linguistic unit. Lingua, 1969,22, 27-46.
Applebee, A. N. The child's concept of story: Ages two to seventeen.Chicago, IL: University of Chicago Press, 1978.
Botvin, G. J. & Sutton-fmith, B. The development of structural complexityin children's fantasy narratives. Developmental Psychology, 1977,13, 377-388.
Bremond, C. Morphology of the French folktale. Semiotica, 1970, 2, 247-276.
Britton, J. et al. The development of writing abilities (11-18). London:Macmillan Education Ltd., 1975.
Britton, J. Language and learning. Harmondsworth, England: Penguin, 1970.
Brunner, J. S. The ontogenesis of speech acts. Journal of Child Language,1975, 2, 1-19.
Carroll, W. S. and Feigenbaum, I. Teaching a second dialect and someimplications for TESOL. TESOL Quarterly, 1967, 1, 31-40.
Cazden, C. Play and metalinguistic awareness: One dimension of languageexperience. The Urban Review, 1974, 7, 28-39.
Clay, M. What did I write? Auckland, New Zealand: Heinemann EducationalBooks Ltd., 1975.
Cook-Gumperz, J. Situated instructions: Language socialization of schoolage children. In S. Ervin-Tripp & C. Mitchell - Kernan (Eds.), Child
Discourse. New York: Academic Press. 1977.
Crystal, D. Child language, learning and linguistics. London: Arnold,
1976.
Davidson, M. L. The multivariate approach to repeated measures (Tech. Rpt.BMDP-75). Los Angeles: University of California at Los Angeles,Department of Biomathematics, Argust, 1980.
DeStefano, J. S. Black English. In J. DeStefano (Ed.), Language, Societyand Education: A Profile of Black English. Worthington, Ohio:Charles A. Jones Publishing Co., 1973.
Dillard, J. L. Black English: Its history and usage in the United States.New York: Random House, 1972.
Donaldson, M. Children's minds. Glasgow: Fontana /Collins, 1978.
Doughty, P., Pearce, J., & Thorton, G. Exploring language. London: Arnold,
1972.
156-127-
4
Dungan, R. Prose memory of good and poor first-grade readers: Effects
of repeated exposures. Unpublished doctoral dissertation, The OhioState University, 1977.
Fasold, R. W. and Wolfram, W. Some linguistic features of Negro dialect.In R. W. Fasold & R. W. Shuy (Eds.), Teaching Standard English in theInner City. Washington, D.C.: Center for Applied Linguistics, 1970.
Favat, A. Child and tale: The origins of interest. (NCTE Research Report
No. 19) Urbana, IL: National Council of Teachers of English, 1977.
Francis, H. Language in childhood: Form and function in language learning.
London: Paul Elek, 1975.
Garvey, C. & McFarlane, P. A measure of standard English proficiency ofinner-city children. American Educational Research Journal, 1970,7, 29-40.
Garvin, P. L. Operation in syntactic analysis. In On Linguistic Method:Selectic Papers. The Hague: Mouton & Co., 1964.
Geisser, S. and Greenhouse, S. An extension of Bex's results on the use ofthe F distribution in multivariate analysis. Annals of MathematicalStatistics, 1958, 29, 885-891.
Graves, D. An examination of the writing processes of seven year oldchildren. Research in the Teachinof English,' 1975, 9, 3 pp. 227-241.
Halliday, M. A. K. Explorations in the functions of language. London:Edward Arnold, 1973.
Halliday, M. A. K. & Hasan, R. Cohesion in English. London: Longman, 1976.
Hasan, R. Code, register and social dialect. In B. Bernstein (Ed.), Class,Codes and Control (Vol. 2). London: Routledge & Kegan Paul, 1973.
Henderson, E. H. The role of skills in teaching reading. Theory IntoPractice, 1977, 16, 348-356.
Hildreth, G. Developmental sequences in name writing. Child Development,1936, 7, 291-302.
Hunt, K. W. Differences in grammatical structures written at three gradelevels, the structures to be analyzed by transformational methods.Tallahassee, FL: Report to the U. S. Office of Education, CooperativeResearch Project No. 1998, 1964.
Hunt, K. W. Grammatical structures written at three grade levels. (NCTE
Research Report No. 3). Champaign, IL: National Council of Teachersof English, 1965.
Jones, L. V. Analysis of variance in its multivariate developments. In
R. B. Cattell (Ed.), Handbook of multivariate experimental psychology.Chicago: Rand McNally, 1966.
King, M., Rentel, V., Pappas, C., Pettegrew, B. and Zutell, J. How Children.
Learn To Write: A Longitudinal Study. Columbus: The Ohio StateUniversity Research Foundation, 1981.
King, M. and Rentel, V. Toward and theory of early writing development.Research in the Teaching of English, 1979, 13, 243-253.
Labov, W. & Cohen, P. Systematic relations of standard and nonstandardrules in the grammar of Negro speakers. Ithaca, New York: ProjectLiteracy Report No. 8, 1967.
Labov, W., Cohen, P., Robins, C. & Lewis, J.English Negro and Puerto Rican speakersPhonological and grammatical analysis.Research Project No. 3288, U. S. Office
A study of the non-standardin New York City. Volume I:Final Report, Cooperativeof Education, 1968.
Langer, S. K. Feeling and form. New York: Scribner's 1953.
Leondar, B. Hatching plots: Genesis of storymaking. In D. Perkins & B.Leondar (Eds.), The arts and cognition. Baltimore: The Johns HopkInsUniversity Press, 1977.
Loban, W. Language development: kindergarten through grade twelve. (NCTE
Research Report Jo. 18). Urbana, IL: National Council of Teachers ofEnglish, 1976.
MacNamara, J. Cognitive basis of language learning in infants. PsychologicalReview, 1972, 79, 1-13.
Mandler, J. M. & Johnson, N. S. Remembrance of things parsed: Story structure
and recall. Cognitive PsychologE, 1977, 9, 111-151.
Maranda, E. and Maranda, K. Structural models in folklore and transformationalessays. The Hague: Mouten, 1971.
Moffett, J. Teaching the universe of discourse. Boston: Houghton Mifflin,
1968.
Nelson, K. How young children represent knowledge of their world in and outof language: A preliminary report. In R. Siegler (Ed.), Children'sthinking--What develops? Hillsdale, NJ: Lawrence Erlbaum Associates,
1978.
O'Donnell, R., Griffin, W. J., & Norris, R. C. Syntax of kindergarten andelementary school children: A transformational analysis. (Research
Report No. q), Urbana, IL: National Council of Teachers of English, 1967.
Olson, D. R. The languages of instruction: The literate bias of schooling.In R. C. Anderson, R. J. Spiro & W. E. Montague (Eds.), Schooling andthe acquisition of knowledge. Hillsdale, NJ: Earlbaum, 1977.
Pettegrew, B. C. Text formation: A comparative study of literate and pre-literate first grade children. Unpublished doctoral dissertation, TheOhio State University, 1981.
153-129-
'111:C1"
Politzer, R. L., Hoover, M. R. & Brown, D. A test of proficiency in Blackstandard and nonstandard speech. TESOL Quarterly, 1974, 8, 27-35.
Preston, E. M. Squawk to the moon, Little Goose. Illustrated by B. Cooney.New York: Viking Press Inc., 1974.
Propp, V. Morphology of the folktale. Translated by Lawrence Scott.Auston: University of Texas Press, 1968.
Read, C. Pre-school children's knowledge of English phonology. HarvardEducational Review, 1971, 23, 17-38.
Rentel, V. & Kennedy, J. Effects of pattern drill on the phonology, syntax,and reading achievement of rural Appalachian children. AmericanEducational Research Journal, 1972, 9, 87-100.
Rubin, S. & Gardner, H. Notes from a presentation at a conference on re-search in literature. Buffalo, New York, 1977.
Shuy, R., Wolfram, R. & Riley, W. K. Field techniques in the urban languagestudy. Washington, D. C.: Center for Applied Linguistics, 1968.
Stein, N. L., & Glenn, C. G. An anal "sis of story comprehension in elementaryschool children. In R. O. Freedle (Ed.), New directions in discourseprocessing. Vol. 2, Norwood, NJ: Albex Publishing Corporation, 1979.
Todorov, T. The two principles of narrative. Diacritics, Fall, 1971, 37-44.
Ure, J. Lexical density and register differentiation. In G. E. Perren &S. L. M. Trim (Eds.), APPlications of linguistics: Selected papers ofthe Second International Congress of Applied Linguistics. (Cambridge1969). London: Cambridge University Press, 1971.
Warner, W. L., Meeker, M., & Ellis, K. Social class in America. New York:Stratford Press, Inc., 1949.
Wheeler, M. E. Untutored acquisition of writing skill. Unpublished doctoraldissertation, Cornell University, 1971:
Winograd, T. A. A framework for understanding discourse. In M. A. Just &P. A. Carpenter (Eds.), Cognitive processes in comprehension. Hillsdale,
NJ: Earlbaum, 1977.
Wolfram, W. A sociolinguistic description of Detroit Negro speech. Wash-ington, D.C.: Center for Applied Linguistics, 1969.
APPENDIX A
Procedures of the Study
The purpose of this study was to describe the transition childrenmake from oral to written texts, in respect to their use of cohesivedevices in two modes of oral, and one mode of written, language, andtheir inclusion of particular story structure elements in the same threemodes. The approach chosen to realize the goals of the investigation wasa longitudinal study of two groups of subjects:
36 children, grade 1 through 2
36 children, kindergarten through grade one
The two populations permitted both, cross-sectional comparisons betweengroups as well as longitudinal comparisons over a period of 16 months.This report, however, will describe only the grade one through giade twopopulation, as required in NLE Grant 7940039. This population wasstratified by sex, school, dialect and socio-economic class. They wereobserved at three-month intervals, across three modes of discourse.:writing, dictation, and story retelling. These three contexts wereexpected to influence the production of texts differentially over thefive observations, yeilding comparisons in the number and kinds ofcohesive ties employed in each mode, as well as comparisons of thestructural characteristics of texts produced in each mode.
Selection of Subjects
To study writing, a first essential was to select schools andclassrooms in which the curriculum encouraged writing from childrenduring the first two years of school. A second necessity was to locateschools where research associates could easily move in and out ofclassrooms to collect data and/or work with individuals or groups ofchildren. A third requirement was to identify schools which reflectedthe characteristics of urban and suburban schools in America including,particularly, the Language and socio-economic differences whie.-. prevailin these schools--because both, language and soctp-economic factors have"been implicated as important factors in school achievement.
The urban school selected as a site for this study contained apopulation of Black children from the neighborhood and a sizeablepopulation of white middle class children transported to the school bybus. This fortuitous situation allowed usl to observe children whosesocial backgrounds differed substantially, and who had in common a newkind of educational environment. Choosing a suburban school allowed usTo7tOmpare-the-mitddle class-children-in-ehe-orban-se-booI-with s-like-population in a different setting. A more detailed description of theschools, hereafter referred to as Urban and Suburban, follows.
161-132-
Urban School
The Urban school, designated as an alternative school, is located inthe central area of a large mid- western city, and it provides schooling
for children pre-kindergarten to grade six. It is an open -space
with multigrade groupings in each work area. The school avoids gradelevel labels and, thus, each large classroom space is referred to as theRed Area, the. Blue Area, or the Yellow Area.
The first year our first-grade subjects were located in the Red andBlue areas and distributed across five teachers. The Red Area housedkindergarten and grade one pupils, and occupied two separate butconnected classrooms. The Blue Area was a vast wall-less carpeted spacethat was open to the library, located a half-flight above. There werethree teachers for the 90 children, two aides, and two special reading
teachers.
The teachers planned jointly and often brought the children togetherfor large-group activities. Most of the work, however, wasindividualized or accomplished through small-group instruction. A verystrong part of the 'program was the opportunity children had to talk withpeers and with adults. The children had the benefit of special teachersin physical education, art, music and drama, as well as the help ofstudents from local colleges, who were at various stages of teacherpreparation.
Because of its location in the dowtown area, Urban used the nearbycommunity resources (e.g., art gallery, Center for Science and Industry,and businesses) as an extension of the classroom. Children in the BlueArea frequently took walking trips' to plaies of interest.
Children from any elementary school in the city may make applicationto attend Urban School. While children in the neighborhood are givenpriority, there is an attempt to make the school population reflect theschool system, as a whole, in terms of racial background, achievement,and socio-economic status.
During the first year of the study, the 24 subjects in Urban weredistributed across five class teachers. The following year they werelocated with six different teachers, and in three work areas:
Teachers CC MB MS DH SB BS6 3 4 7 2 6
This distribution, of course, made observations and work withchildren extremely time consuming and data collection very complex.While teachers were similar in their concern for children and theirlearning, they differed greatly in teaching style, approaches -to-'Ateracy, and interest in children's writing. They'were not expected tofollow a set course of study in reading and writing, but rather, were inthe process of developing one for their school. While this gave theteachers and children a great deal of freedom, it meant that thecurriculum was ever chinging and not very predictable. Emphasis ia
-133-
162
literacy instruction was on skills--in word recognition, handwriting, andspelling. A wide range of textbooks, audiotapes, and duplicatedmaterials were used in teaching reading--usually at the discretion ofeach teacher. For instance, one teacher used experience stories writtenon charts, as a means of teaching reading.
Over the 15 months of the Project, change in emphasis and materialsdid occur. More attention was given to the content of children'swriting, to exposing children to clusters of books and stories of asimilar genre, and to reading aloud to children and telling stories.
Suburban School
The Suburban School was located in the oldest part of the mostaffluent suburb in the metropolitan 4%rea. It too was an alternativeschool for parents in that city who wanted their children.to be educatedin an environment that was less formal and prescriptive than thatexisting in most schools in the district. The school, which served apopulation of kindergarten through grade six, was housed in threeseparate buildings or "pods," each consisting of four classrooms. Theschool was locatedon the same grounds as the oldest elementary school inthe district. Some facilities (library, playground, gymnasium) andresources (special teachers and health services) were shared, but theadminiatration and curriculum were separate.
For almost a decade a core of teachers and the principal of theSuburban School had been studying and implementirg informal orprogressive approaches to educating children. The classrooms werearranged with work areas, including resource centers with materials forart, mathematics, and science; book and quiet reading area:; and openspaces where the class couli meet as a group. Most Instruction was.individualized or conducted in small groups. The children were free tomove about the classroom and to work with nne or two f: ',ends; thus, peerteaching /learning became an important elk. At in the instructionalprocess. Every effort was made to integrate the currIzulum which wasorganized around focal interests or longer units of study. The firstgrade, for example, typically studied foods and visited a super - marketand distribution center. The second /third grade class pursued interestsin witches, horse?. plar*s, and the human body. Reading and writing wereusually integrated with these projects, but some small group andindividual instruction was given to reading. A great emphasis was placedon literature and using a range of books, both fiction and nonfiction, inall studies. Literature was studied or itself too. Teachers frequentlyread aloud to children, discussed books with them, and often organizedbooks for study around a common theme, concept, author, or illustrator.
The teachers varied, of-course, in their understanding of integratedlearning and ability to implement the concept. They varied also in theirbeliefs about effective ways to foster literacy. When the WritingProject began, the subjects in mid-first grade were distributed acrosstwo classrooms: one was a kindergarten/first gradc; the other, afirst/second grade, In both classes, teachers used a modified Aanguageexperience approach in which experiences were charted. In turn, these
163-134-
charts often were copied by children. Great emphasis was placed oncorrect spelling and capitalization, so lists of words in manuscriptwriting were made available to children before they began any personalwriting. This emphasis changed over time as teachers saw that childrenhad more spelling ability than they had been able to use and that theywrote more and better texts when freed from spelling constraints.
The second year of the study the subjects were again distributedover two classrooms, both containing pupils in grades_two and three.Again, the teachers differed. One placed strong emphasis on language andliterature, and the other emphasized science and physical activities.Both, however, participated enthusiastically in the study and appreciatedthe growth in writing they saw their children experiencing.
Subjects (24) were drawn from the first grade of an "alternative"school, an elementary ..zhool so designated because of its openenrollment, open- space, and open curriculum:" nits school was atended-by
11children not only from a largely Black neighborhood with an SSSdistribution ranging from low to lower middle class, but also from middleclasc neighborhOods throughout the city. An additional rAmpla (12) wasdrawn from the first grade of a suburban school with a Socio-economic
21.67; SD =a 5.99). Subjects scoring ten or
Status (SU) distribution ranging from middle to upper class. Fron theformer population, 12 subjects were identified as vernacular Blackdialect speakers, using the revised measure of standard Englishproficiency noted above (M asmore on this measure were assumed to be vernacular Black dialect
identifying Black-Vernacular Speakers
I
speakers.
We hypothesized that dialects or codes may be related to exophoricreference. evidence suggests that speakers of Black English varyconsiderably, both as individuals, and as a group, in the number and
11kinds of forms they produce in varying circumstances (Carroll andFeigenbaum, 1967; DeStefano, 1973; Dillard, 1972; Labov and Cohen, 1967).
Tp assure that subjects spoke vernacular Black English, threealternative screening techniques were considered: (1) technical detailedlinguistic interviews (Labov, Cohen, kobins, and Lewis, 1968; Fasold andWolfram, 1970); (2) semi-informal interviews (Shuy, Wolfram, and Riley,
111968); and (3) sentence repetition tasks (Garvey and McFarlane, 1970;Politzers Hoover, and Brown, 1974; Rentel and Kennedy, 1972). Given theinter- and intra-subject variability noted above, sentence repetitiontasks were employed because these tasks discriminate among subgroups onitems where a difference exists between the form presented, and a formhabitually used by- a subject and offered as a 3ubstit 'e,_ with relatively_ 111_
high reliability (Garvey and McFarlane, 1970). In as ition to theadvantages of increased discriiinability and reliability, sentencerepetition tests require less time and less exacting training for theirproper administration. Ten structures from the Garvey and McFarlane II"
164
scale with reliability coefficients greater than .55 were selected andincluded in the scale, (four repetitions of each structure) for a totalof 40 items (see Appendix A).
Determining Socio-Economic Status
During the first few weeks of the study (February 1979), thesocio-economic status of those children for whom parental permissionforms were received was determined by using a modification of the Indexof Status Characteristics (Warner, Meeker, and Ellis, 1949), a scalewhich rates occupation, source of income, house type and dwelling area(see Appendix 8). Because Warner's occupation ratings are dated,Hollingshead's Job Scale was substituted and weightings adjusted.Weighted totals of the four subscales comprised the SES score for eachsubject. The total scale had a range of 12-84.
All 2C of the vernacular speakers fell within the bottom quartile ofthe SES distribution, leading to the conclusion that, at least withinthis population* their dialect was socially constrained--that is, asocioUct (DeStefano, 1973). From this population, six males and femaleswere drawn at random (M = 71.00; SD = 8.51). Middle class subjects weredrawn from both, the same inner city school, and from a suburban school,(six males and six fec4les from each) iu order ro contrast school andcontrol for class differences.
Dugan (1977) found that first-grade t4ys differ-4 significantly fr..=first-grade girls both, in the amount, and kinds of information theyincorporated into their retellings of stories. Sex also appears to be afactor in the number of vernacular black forms produced by a speaker(Wolfram, 1969), women using fewer Black English forms than their mLle,ghetto counterparts. To control for these expected differences, sex wasincorporated into the design of the study as a blocking variable.
One of the most vexing problems Ln longitudinal research is, ofcourse, subject mortality. To compensate for the possible loss ofsubjects from the group of 36, initially drawn at random from the totalstratified subject pool, two additional subjects were drawn randomly &aneach level of the pool--as not t iearlier, stratified bydialect/socio-economic class, sex and schooland assigned to each levelof the design. Data were obtained chew 12 replacement subjects, allblind to their identi f as replacements. Thus, eight subjects wereassigned to each celL constituting the blocking variables in the study.Two subjects were lost from the lower class, female, vernacular-speaking,urban-school cell. Two also wers lost from the middle class, female,nonvernacular-speaking, urban-school cell. To obtain equal numberswithin each cell, two subjects were dropped at random from the remainingfour cells in the design for a total of 36 subjects.
To determine the extent to which the assignment of replacementsuUject* to the design had affected the composition of these levels,scores for middle class subjects from the Index of Status Characteristics
were subjected to an analysis of variance having two between-subjectcomparisons--sex and school. The results of this analysis are presentedin Table 1.
Table 1
ANOVA of Socio-Economic Class by School and Sex
Source df MS F g <
School (A) 1 222.04 5.39 .05
Sex (B) 1 35.04 .05
School I Sex (A x B) 1 22.05 .54Error- (W /Ss) 20 41.19
Total 23 47.95
As can be seen from Table 2, subjects from the suburban schoolscored significantly lower on the Index of Status Characteristic4. As
indicated by Table 1, there were no other significant effects.
Table 2
Means and Standard Deviations of Socio-Economic Class by School and Sex
Index of StatusCharacteristics
Urban SuburbanSchool School
MeanStandard Deviation
38.33 32.257.47 4.41
Quite obviously, replacing subjects in the uroan school populationunbalanced the equality that had been established within the middle classpopulation for the two schools. This finding of school differences, thusnecessitated a design arrangement wherein the suburban population had tobe treated as a distinct subgroup. Therefore, data from the suburbanschool were analyzed, both separately, and in a school replicationarrangement for all MANOVAS, ANOVAS, and discriminant function analyses.These design arrangements are discussed in later sections of thischapter.
16d
The import of this difference between the urban and suburban middleclass populations must be kept in perspective. The Index of StatusCharacteristics, the socio-economic scale emplcyed in this study, has aweighted score range of 12 to 84. Both means reported in Table 2 restwell below the midpoint of the scale (48), and clearly within the "middleclass" spectrum on the scale. Whether or nor treating class extremessuch as "middle class" or "lower class" has any greater import forlanguage variation than significant differences found to exist withinthese larger categories has not been established. But, there is no goodreason for ignoring such "within-class" variations. Therefore, thefinding that middle class children in the two schools differedsignificantly on the Index of Status Characteristics argued for theinclusion of a school replication study as a minimum and separateanalyses for each school, as necessary, where differences in thereplication study were obtained.
Data Collection Procedures
During the early weeks of the study, research associates worked inthe classrooms with individuals and small groups of children. They readstories to them, invited children to retell the stories, or to tellothers they knew." The research associates also encouraged them towrite, often providing materials in the form of colored paper, booklets,or flow pens. Children also viz:re given the opportunity to dictatestories of their own composition, with the researcher acting as scribe.The oral story retellings, as well as the dictated stories, frequentlywere audiotape recorded to prepare the way for the recording to be doneas a part of the later data collection. These activities were carriedout in the regular classroom or other available vacant rooms in theschools. Prior to the actual data collection, all children had theopportunity to hear, tell, and dictate stories.
The language samples in the three modes were collected in March1979, October 1979 and May 1980. Seven research associates participatedin the data collection, but all had been working in the classrooms andwere known to the children as visiting teachers. At least one associateworked regularly with each classroom and knew the children well. Allresearchers were trained to data collection procedures (see appe-.41ces Cand D).
Story retelling data usually were collected in a single day atsenool, this was followed by the collection of dictation data, whichrequired three or four days in each school. Every effort was made to fitthe dictation and writing experiences into the or.going life of theclassroom. The writing was carried out in the classrooms, with theteachers discussing the assignment with their children.
Story Retelling
Three very different folktales were chosen for the retellingexperience. The quality of the story, reasonable length for retelling,and children's lack of knowledge of the tale, were among the criteriathat influenced selection (see page 35 for others).
In small groups of four to six, children were taken out of theclassroom to a room in the school where the stimulus story could be readwithout interruption. One member of the research team served as story"reader" and the others as "listeners" for the retellings. The childrenwere told they would be read a new story that the reader had enjoyed andwanted to share with them. The reader also told the children they wouldeach have the opportunity to share the story with a visiting teacher whenthe reading was finished. The reader then read the story as it typicallywould be read in the classroom, providing enough time so that thepictures could be viewed. Upon completion of the story, the reader wentthrough the book a second time, showing each page in turn, not commentingbut accepting any spontaneous comments about the story from the children.If, at any time, a child indicated concern about being able to remembereverything about the story, in retelling it to another, he was reassuredthat it was all right to retell only what he could remember.
Following the reading each child was taken to a "listener" member ofthe research team who was introduced as a teacher who did not know thestory that had just been read. The number of listeners matched thenumber of children in each story reading group so that no child was madeto wait, i.e., the time and activity between the end of the readingsession and the retelling was uniform for each child. In introducing thelistener, the reader explained to each child that the visiting teacherdid not know the story that had just been read and stated that theteacher would like to hear it The reader then left the room, thelistener reaffirmed the task, explaining that the retelling would be taperecorded for the purpose of sharing i1. with other teachers who wereinterested in stories. Once the child began his retelling, the listenertried not to interrupt the child's narrative. The listener wasattentive, but did not collaborate in the child's text production. Theintent 'was to allow the child to construct his own text and to avoidadditions by the listener to the content or structure of the narrative.
Dictated Story Data.
Dictated stories were collected at the two schools during thetwoweek data collection period, exclusive of the two days devoted tostory retelling. Expectations for sic rating original stories to membersof the research team had been established prior to the data collection;all children had previous experience in dictating stories to a researcherwho acted as scribe while being tape recorded as an ongoing classroomactivity. The child was told that his story was to be written for him,
16
that it could be as long (or as short) as he wanted, and that it could beabout anything that interested him/her. Emphasis was placed on composing"your very own story," rather than retelling a well-known one (e.g., TheThree Little Pigs") or a recently-viewed TV cartoon.
The story was taken down in manuscript writing by the researcher.The child was aware that his words were being written and could see theactual writing if s/he wished. Children were given an unobstructed viewof the scribe's activities and ample opportunity to observe the scribetake down their dictations.
Dictation proved to be a fairly popular activity in first grade,with most children requesting a turn with the scribe. Generally theorder of data collection followed a volunteer pattern, with the scribesworking with children who indicated their readiness with a story. At thetime of collection each child want with a scribe to an available room inthe school where a tape recorder had been set up. The dictation sessionwas tape recorded, and the child was told that the purpose of therecording was to check on the accuracy of the scribe's copy before it wastyped and placed in the classroom storybook. Once the child begandictating, the researcher attempted to keep up with the child's dictationpace, accepting any comments or instructions the child gave regarding thescribe's performance and/or the writing process, but was careful not tointerrupt the child's narrative. In cases when a child dictated anobvious retelling of a known story or rhyme the scribe elicited a seconddict..-Led text after encouraging the child to till his/her own story (seeDictation Procedures, Appendix D).
Story Writing Procedures
During the two-week observation period, an "assigned writing" samplewas collected from each subject. Every effort was made to make thisactivity a natural part of the ongoing work of the classrooms. But insome situations, particularly in the early collections in grade one, thechildren were not accustomed to writing original stories. In fact, manydid very little writing, and what was produced often was copied fromcharts or the chalkboard. In the beginning, it was therefore necessaryto develop, with the teachers, conditions that would interest childrenand cause them to write a story within a paiiod of one or two days.Emphasis was placed on writing stories. Thus, children were givencolored paper or paper folded into booklets to further establish thestory context. Teachers discussed the writing assignment with thechildren and tried to link it to work and experiences that children werecurrently involved in. Sometimes the discussion centered about stories,a wordless picture book, or a recent particular experience a visit to agrocery store, or a performance by a mime. The contexts were varied, buta first priority of the investigations was to work within the curriculumand constraints of each classroom.
163-140-
Sessions for assigned writing were not limited in time.Nevertheless, the children normally were to begin in early morning andcontinue for an hour or more, or until most children were finished.Anyone who had not finished and wished to do so, kept his story to workon through the afternoon and next day. The researcher, as well as theclassroom teacher, was available in the initial writing session. The
researcher then returned the next day to sit down with the authors andread through the stories. This last step was essential because childrenwere encouraged to use their personal, creative, or invented spellings.Occasionally these renditions were beyond interpretation without the helpof the author. The exact word intended was essential for the cohesionand story structure analysis, as well as for the spelling coding.
As soon as the writing was obtained, two copies were made and theoriginal returned to the classroom, if so requested by the teacher. In
most instances, however, the original script was retained.
Preparing the Oral and Written Texts for Coding
Preparation of the transcriptions of the audiotaped oral narrativesproduced in the two tasks (story retelling and story dictation) proceededin two stages.
In the first stage, a complete transcription of each audiotaped datacollection session was made. The stream of speech was initiallysegmented at the level of the orthographically realized word.Transcriptions were typed in traditional orthography with capitalizationof proper nouns and the first-person singular pronoun. No punctuationwas included in the typescripts. These original typescripts wereunedited and included all verbalizations recorded during the sessions.Filled pauses, word and phrase repetitions, stutters, corrections andfalse starts were included, as were any verbal interactions between childand listener/scribe. Interjections by the adult were rare, but when theydid occur, interjections typically consisted of indications of continuedinterest such as "him" or repetition of the child's most recent wordsfollowing an extended pause. Unintelligible words or segments of text,which occurred very rarely, were noted in the following manner on thetypescripts: ( ), for what appeared to be a single word, and( ), for longer utterances. Lines of typed text were numberedsequentially and words spoken by the listener/scribe were identified withthe letters: IN. (An example of an original typescript appears inAppendixi.)
Using both the prepared typescripts and the audiotapes, a researchassociate, working with a second researcher, edited the typescripts inpreparation for coding. First, each child's narrative text wasabstracted from the total language recorded during the taping sessions.There was no difficulty in determing text boundaries; the two editorsagreed in all cases. Context supported by the children's use of
narrative conventions such as "once upon a time..." or "there wasonce..." and "they lived happily ..kver after," facilitated boundary
decisions. Also of help in many language samples, was a shift into a"story voice" distinct from the conversational language intonationpreceding and following the narrative text. Marked for exclusion fromthe analysis were non-silent phenomena such as filled pauses, unmotivatedrepetitions, and abandoned forms. These non-gilent phenomen4 correspondto what have been called "mazes" (Loban, 1963), or "garbles" (Hunt,1964), in descriptions of child language. Editors also ma:kedlistener/scribe interjections and child asides (examples of the latter:Nj wanted 'landed"; "did I say 'pigs'?"; you like writing, don't your)for exclusion from the narrative texts. Examples of verbalizationsexcluded from the narrative texts (marked by brackets and asterisks) aregiven below. The first example is from the retelling corpus and thesecond. is from the dictation corpus.
(2.11 once there was an old woman and her littlegirl and they were really poor and they onlyhad (a little' a tiny loaf of bread and thenevery day the little girl would go out (tofind) to the woods to find some nuts andberries ... #
(2.21 (um] the witch (um] went to feed thebogs then luml the witch went to feed thechickens then the horses* did I say pigsdid I say pigs*INs**you said hogs***oh then pigs* (she went to feed] she wentto feed the pigs
Editing also involved identifying and marking the units upon whichthe subsequent cohesion and story structure analyses were to be based.While cohesion, Halliday and Hasan (1976) point out, is not limited torelations "above the sentence," the present study focused on the meanswhereby structurally unrelated units of language are linked together.Halliday and Hasan refer to this "intersentence cohesion" as "thevariable aspect of cohesion" (1976, p. 9). The analysis of"non-structural" cohesion requires the identification of sentences orsentence-like units in the language to be analyzed. Linguists point outthe difficulty of defining the "sentence" (Allerton, 1969; Crystal, 1976;Garvin, 1964). As Allerton notes, traditional definitions of thesentence are made in terms of the conventionalized written language,i.e., as a sequence of words lying between punctuation marks. Suchtraditional definitions were not useful for the oral language data ofthis study; therefore, an operational definition of a sentence-like unitthat ccula deal with spoken English was selected: the "T-unit." As:leaned by Hunt (1964), the T-unit is a complex clause consisting of oneindependent or main clause with any dependent or subordinate clausesattached to it of embedded in it. The T-unit has been used in manystudies of child language development--in both speech and writing--
174
because of its efficacy and reliability. This kind of reliability isparticularly important to the present study of the cohesive relationsbetween non-structurally related elements of children's oral narratives.
An additional editing procedure involved segmenting, or parsing, thetexts into the T-units, upon which the cohesion analysis was based. Alsoat this point, selected symbols, found to be helpful during cohesionanalysis in interpreting text and making coding decisions, were added tothe typescripts. The full notational system used in editing thetypescripts is presented in Figure 1. And an example of an editedoriginal typescript appears in Appendix F. Following the editingprocedure, typescripts were retyped, and coded identification numberreplaced all other identification on the protocols.
One copy of the children's writing was kept in its original statefor analyses related to concept of message, spelling, and other writingconventions. The second copy of all those scripts judged to be a textwere cast into T-units, edited, and transcribed (with all spellingscorrect), following the procedures used for the oral '..exts. Storystructure and cohesion coding were done on the typed scripts that hadbeen parsed into T-units.
Cohesion Coding and Analyses
Coding of the edited narrative texts followed the scheme set out inCohesion in English (Halliday and Hasan, 1976). The five categoriesidentified by Halliday and Hasan which represent types of cohesion(reference, substitution, ellipsis, conjunction, and lexical cohesion),provided the framework ior coding. Ail instances of exophoric, as wellas endophoric, presupposition, within these categories, were coded.While not contributing directly to the integration. of a text (i.e.,cohesion, as technically define' , exophora does contribute to thecreation of text through linking language with features of the largertextual enraonment and, as such, bears on the question of interest inthis study; what options do (hildren use in creating their texts? Allcoding was done by two research associates and one principalinvestigator. A reliability check was run on a sample of tenrandomly-selected texts, five representing each task. A researchassociate trained in cohesion analysis also coded the ten texts. Thecorrelation coefficient calculated for the two coders was .96 (SPSSSubprogram Reliability).
As noted earlier, exophora is a type of phoricity which takes oneoutside the text. Exophoric items are presupposing textual elements,whose intended, more precise meanings, are mediated throughextra-linguistic factors. !chile it is possible for the presuppositioninvolved in reference, substitution, and ellipsis to be exophoric,occurrences in the latter two categories are fairly infrequent (Hallidayand Hasan, 1976).
A -143-
1'72
Used to mark the boundaries of each narrative text.
Used co mark non-silent phenomena (filled pauses,unmotivated repetitions, abandoned forms, etc.)and, following Hunt, considered extraneous to theT-uait.
This mark identifies listener/scribe interjectionsor child asides not considered a part of the child'sintended narrative text.
Used to mark any responses to interjections or asidesnot considered a part of the narrative text.
Slashes mark T-unit boundaries and are numberedsequentially.
1 Question and exclamation marks were added to thetypescript when the child's intonation warrantedit and proved helpful in subsequent cohesionanalysis (no other terminal punctuation was marked).
Quoted speech in the text for which a speaker islexically identified.
((sp:name))"..." Quoted speech in the text which is not lexicallyattributed to a sftaker but which can beattributed to a speaker based on context or thechild's use of a role voice.
((sp0)) "..." Quoted speech in the text which is ambiguous withrespect to speaker.
underlining Underlining is used to mark contrastive stress orother kinds of emphasis used by the child whichcould aid the cohesion coder in interpretating thetext.
Figure 1: Notational System for EditingOral Language Transcripts
173-144-
A system for subcategcrizing exophoric reference was adapted fromBasan's forthcoming work (in press) on semantic styles. The
subcategorization is based on the type of situational knowledge requiredfor interpretation of the exophoric item. Using the criteria andterminology proposed by Hasan, the following subcategories of exophoricreference were coded its the data of this study:
Formal Exphora--Those items which are only technically exophoric.One's knowledge of the language and a shared cultural contextallow an adequate interpretation. Thus, upon hearing orreading the utterance, "On her way home from school thereluctant scholar dropped her books in the street," one doesnot feel compelled to identify what street. Specificidentification of the entity marked by the definite articleis, in this instance, irrelevant. "Generalized" exophoricreference ("You (i.e., one) shouldn't feed the animals atthe zoo"), "institutionalized" exophora ("Jim went to see thepolice"), and "homophota" (reference to a whole class or to aunique member of a class, such as the stars, the moon) wereincluded in this category.
Instantial Exophora--Those items whose presuppositions are mediatedvia some elements in the immediate situation: reference isbeing made to some aspect of the here-and-now. For example,if an author begins his story with, "I went to Mars on aspaceship and had a great adventure," full identification ofthe referent of the pronoun is situationally possible. Evenif not present at the text's creation, a partial identificationof "author" is possible and usually adequate. In the narrativetexts of this study, instantial exophorics were limited tofirst- and second-person proncouns.
Restricted Exophora--Those items whose irmended meanings gocompletely beyond the immediate sio,ltion and are availableto the listener/reader only on the 'oasis of shared knowledgemediated by past experience. Thus, in a story retelling thatbegins, "They didn't have any food--just this little piece ofbread. She went out to look for nuts and berries,"identification of "they" and "she" is not possible withoutrecourse to knowledge that goes beyond this retelling situa-tion and this text. (If the illustrated story on which theretelling is based were present during the retelling, and thepictures were pointed to, then these exophora would beconsidered instantial. The book, with its illustrations, wasnot available to the child during the retelling task in thisstudy.)
The semantic constraints involved in telling a story to another whoclaims not to know the story, require that one talk in such a way thatone's meanings are available to the listener. The use of formal exophoraand certain instantial exophora (Chose representing speech roles in the
situation) in the tasks of this study, were seen as unambiguous in these
contexts of narration. However, the use of restricted exophora relativeto the characters and events in the stories, was seen as ambiguous. Inthis study, formal and instantial exophora, whose meanings wereconsidered available to the listener, were included for purposes of dataanalysis in the category of endophoric reference. Restricted exophoraformed a spearate category for tabulating purposes. Thus frequencieswithin six categories of presuppositional "ties'' were tabulated:reference, restricted exophoric reference, ellipsis, substitution,conjunction, and lexical cohesion. Appendix G contains an edited,retyped dictation text, along with a sample of the coding record for this
text.
Analysis of the Cohesion Data
Differential use of cohesive ties in writing was compared in threeseparate MANOVAs where dialect, school, and sex served as thebetween-subjects factors and Observation analyses, the within - subjects
factor. MANOVA (Jones, 1966) was selected because it permits the testingof group differences in terms of multiple dependent variables consideredsimultaneously. MANOVA packages the dependent variables into atransformed composite variable, Y, which represents a linear combinationof the response variables weighted to maximize a discriminant criterion.A significant MANOVA test statistic suggests :ejection of the nullhypothesis of no difference among group centroids. If overalldifferences among groups are found, follow-up techniques allow theassessment of the relative contribution of each of the dependentvariables to those differences.
Three separate comparisons were made because, in each instance,there was no comparable population. In one comparison, the objective wasto explore differences between schools; in another, differences betweendialects within a single school; and in the third, differences betweensex over observations. They are listed below:
MANOVA 1
MANOVA 2
MANOVA 3
School X Sex X Observation
Dialect X Sex X Observation
Sex X Observation
Figure 2. Cohesion Multivariate Analyses of Variance
Text length was free to vary in the narrative tasks of this study.To allow for differential text length, frequencies of ties within the sixcategories identified for coding were expressed as a proportion of totalties for each text. Following the coding, it was observed thatreference, conjunction; and lexical cohesion were used extensively by allchildren in the tasks. Ellipsis and restricted exophoric reference were
175-146-
used by most of the children. Moreover, use of these latter twocategories of linguistic devices involved more than one instance in thegreat majority of cases, although their relative frequency of use did notapproach the magnitude found for reference, conjunction, and lexicalcohesion. Substitution,' however, as a text forming device, was used byfew children in the samples, and even fewer had more than one instance ofsubstitution in their texts. Therefore, this category was eliminatedfrom the multivariate analysis of variance, performed on the proportionscores of the remaining five categories. These categories were:exophoric restricted reference, ellipsis, conjunction and lexicalcohesion.,
Since proportion scores were to be used in the MANOVA, the were
subjected to an arcsine transformation to conform to the assumptions ofthe multivariate normal distribution. The accsine transformation resultsin a variable that is normally distributed with a constant variance.Computer program CANOVA, a component analysis of variance (ClydeComputing Services, 1973) was used for the MANOVA analysis. The test ofsignificance employed was Wilks's likelihood ratio criterion, transformedinto Rao's apptoximate F.
ptory Structure Coding and Analyses
Texts may be thought of as having fixed and variable elements. )Thepurpose of text analysis is to characterize these two properties. .Propp(1968) attempted to specify the fixed properties of Russian fairy talesaccording to the functions of the dramatis personna, focusing upon whatcharacters do rather Than upon who carries out actions or upon howactions are acconplished. Functions abstractly represeqt actions. Theyare defined without reference to the character who performs them. Aperson who helps the hero satisfy a need can vary from tale to tale. Thehelper can be a witch, the hero's friend, or a stranger. The underlyingaction .is the same. But since the action does sake place within theoverall set of actions that go to make up the tale, a given act can havedifferent meanings. Someone who helps the hero obtain an agent necessaryfor satisfying a need renders a service far different from a person whohelps lure the hero into a trap. Thus identical acts can represent quitedifferent functions. And quite different acts may have the same meaning.For example, a warning to a child not to go into the forest differasignificantly from one given to a combatant in the course of a conflict.A function is always defined relative to its significance for the courseof the action.
Functions, therefore, serve as fixed elements in a tale. They are
the basic constituents of the story. Propp identified 31 functions. Notall functions, however, must occur in a single tale. When functions dooccur in a tale, they ordinarily do so in a particular order. Thus,
order constitutes a second fixed element in a tale. Order grows out ofthe elemental logic of actions. Help cannot be given without somepre-existing need for it or without some circumstance
176
-147-
erein the hero's plight is made obvious. Likewise, the transfer ofmoney must be preceded by a clear need or a rendered service. Thus,order derives not from convention, but from the logic of events and
actions. Tales with the same functions and orders are most likelyrepresentative of the same genre. But too much should not be made of
order. Even in Propplls analysis of Russian tales, he was forced to positthe notion of transformations to ..-count for tales whose functionsappeared in a noncannonical order. If the order of functions followslogically from the nature of the actions, then it is not necessary topreserve cannonical order.
Subsidiary or minor tale's may be embedded within, or follow upon,
the major tale. Propp referred to these subsidiary tales as moves. The
terutnology is not critical. Thus, we toe referred to all suchsubsidiary actions as moves. What is significant about tLem is thatparallel, repeated, and sequential moves, complicate a tale, giving riseto the question of how -uch subsidiary moves are to be coded and 'tcaled.Propp, of course, solved the problem by breicket:73 moves. He specifiedthat two functions were the basis for assigning a brackez, i.e., villainy,and lack. In addition, two pairingsstruggle, coupled with victory, anda difficult task, coupled with its solution--constitute mutuallyexclusive elements, distinguishing "any tales from seeker tales. Atale, conceivab. could contain both pairs, one pair, or neither ,air.Their presence simply helps to distinguish between moves, but in no wayshould be considered obligatory. What is obligatory is villainy. or lack.
Functions may have double meanings. For example, in Magic Porridge Pot,the mother lacks knowledge of the witch's interdiction, which, of course,she cannot help but violate. Both lack and violation of an -erdiction_
were coded becau: . both meanings were inherent in tne action 4:at ensued.A text also may be vague in terms of the actions of a character wi zn, inturn, makes functions difficult to assign. For example, the text says:
"Mother Goose was going out." But no further mention is made of heractions. Is Ois sufficient as a case of absention? Coding in theseinstr.aces was governed by the principle of assigning functions on thebasis of consequences. Did the tale proceed as if absention occurred?If so, then the meaning of the function was absention and so coded. If
the tale continued with subsequent actions .adicating Mother Goose didnot go out, then absention was not coded. questions of this sort werealways resolved by defining the funetior according to its consequences.
Interjadge Reliabilities for Coding_Proppian Functions
On separate occasions, the same pair of judges coded two sets ofproto,- is from two different story retellings. Interjudge reliabilities'were computed for each set of 20 protocols (.85, .89). Dictationprotocols (36) were coded by a different pair of judges, who achieved aslightly higher level of reliability than the first pair (.9". Overall,however, reliabilities were sufficiently high to warrant confidence infunction definitions and coding procedures (see Appendix H).
177-148-
Genre Classification
After
extablishediscourse.specifiedgenres of
1.
judges had been trained and interjudge reliabilities had been
d, each protocol then was classified as to its genre ofFor, even though task instructions to the children had
that they tell or write stories, many children produc...d other
text. Protocols, thus, were classified as follows:
No Text--No utterance produced by the child.
2. Statement /Label - -A single word or phrase defining or4escribing something in the immediate environment. Forexample, "/t was a duck," or "Desk."
3. Composition --A present tense depiction of a child': currentexperience. Compositions are closely identified with thecircumstances, in and for which, they are produced, i.e.,completing a writing assignment for the teacher. Toillustrate: "My mom is nice. I go to school. My mom lovesme."
4. Interaction - -A text with many elements of a dialogue havingan implied listener with whom an experience is being shared.For example: "First, you draw a circl.e. Than you draw a line.Then you make another line here."
5. Chronicle--Narrative that parallels real events in a child'slife, yet expressed in a story frame with conventions such as,"Once a little girl and boy vent to Disneylmd." Charactet3and actions that parallel non-fictive experience and thematilunity, characterize these texts.
6. Tale Narrative that sets forth events and circumstances thatmay reflect real life but without essential dependence onhistorical fact. They have thematic unity, conventional storymarkers, and fantastic characters, as well as fantastic events.They are fictive in nature.
Following genre classification, chronicles and tales were coded andscored for Proppian functions by five judges blind to subject identitybut aware of context variations. There was no way to conceal *4)esedifferences entirely, because retellings, of cours , were about tte lamewell known stories. Only retellings and dictations were compared.Despit' instructions to the contrary, many children failed to producecnrnnicles and tales in the writing context, thus precluding comparisclswith a measure that presumed a story genre. As reported above,interjudge reliabilities were moderately high. Still, occasioeal codingproblems and questions arose. Two judges resolved such questions andaesigned a function as agreed. It should be noted that in scoring theretellings, no attempt was made to assess recall. Only the functions
178-149-
found in the children's texts were scored, regardless of whether or not acounterpart for a given function could be found in the tale the childrenhad heard. The present study sought only to compare "packaging" andproduction of functions. Studies of the role of memory and comprehen3ionin production are under consideration for later analyses, and onecompleted study will be presented in chapter 6.
Selectin& Stories for Retellings
In selecting stories for retelling, a main concern was to findstories that were not known to our subjects, but would likely interestthen. Our subjects varied greatly in their experiences with traditionalliteratcre. They ranged from one group, that seemed to have someacqumintence with almost all stories considered, to another whosebackgrounds were meagre. Selecting stories became more of a problem thanoriginally anticipated.
At the onset of the proj.ec, most Russian fairy tales were too longand complex for some of our subjects. We looked for well-formed andartfully illustrated folktales, especially for recently publisfted ones ornew versions of old tales. To heighten interest, we caose to use picturebooks, but this decision constrained our choice of stories.
Three very different stories were eventually selected for storyretelling--a modern fable, a folktale, and a Russian fairy tale.
Squawk to the Moon, Little Goose, by Edna Mitchell Preston,illustrated by Barbara Cooney Viking, 1974).
Magic Porridge Pot, by Paul Caldone (The Seabury Press, 1976).
Salt, by Harve Zemach. illustrated by Margo temach (Holt, Rinehartand Winston, 1967).
Squawk to the Moon, Little Goose is a story of lack that has,embedded within it, three brief tales of villainy which provide thetrebbling element found in many fotl:.tales. The story also containsfolktale features of trickery, and also rifrain, as with Little Goose's,"Good's good and bad's bad."
In Puoppian analysis, the tale had two moves.
a (beginning situation) '
2 (interdiction) coupled with 1 (absentation)8a (lack: maturity and insight) and 3 (violation of interdiction)
6 (trickery) coupled with 7 (complicity)$b (villainy)
10 (counteraction)11 (departure)12 (preparation)
173
-150-
13 (reaction)15 (translocation)8b (villainy)
9 (mediation)
10 (counteraction) coupled with 14 \receipt)
18 (victory) coupled with 19 (liquidation)
20 (return)
30 (punishmen031 (equilibrium)
Magic Porridge Pot is one version of the magic pot tales that existin sevut..1 different cultures. It is especially appealing to childrenbecluse it is the mother sho uses the magic pot 'without permission and asa :esult creates a huge problem which the daughter solves.
Actually, Magic'Porridge Pot is two tales, conjoined by aninterdiction given in the first, and violation of the interdiction, inthe second. In Propp's terms, it is a tale with two moves:
a (beginning situation)8 (lack) joined with 11 (departure)9 (mediation)12 (function of donor) and 2 (interdiction)14 (receipt of magic agent) and 15 (transference)19 (lack liquidated) and 31 (equilibriva)
The final state of happiness in the first tale provides the beginning forthe second.
1 (absention)8a (lack) and 3 .violation of interdiction)20 (return)19 (lack liquidated)31 (equilibrium)
Salt is a story of the younger brother, "the fool," succeeding inmaking hii fortune while his two older brothers turn to villainy and
fail. It is a tale of lack--lack of status, success--in which a tale ofvillainy is embedded. The villainy tale is interrupted a giant'sstory, a tale of interdiction and lack.
a (beginning situation)8a (lack), 11 (departure) and 12 (donor)14 (magic agent) 15 (transference)25 (difficult task) and (solution of task)30 (reward to hero) and .11 (promise of marriage)
a (beginning situation) and 11 (departure)5 (delivery of victim to villain) 8 (villainy)
a (beginning, giant's tale)(lack of transport, giant's lack of happiness)
2 (interdiction) 15 (transference)20 (return home)
27 (recognition of hero)
2$ (exposure of false hero)30 (villainy punished) 31 (wedding)
3 (interdiction violated)25 (difficult task)
26 (solution)31 (equilibrium)
These stories were analyzed to determine their comparability il.terms of Propp's functions. The criteria on which they were comparedwere; (a) total number of functions in a story, (b) the number ofdifferent types of functions in a story, and (c) the number of moves in astory. As noted earlier, a given function may occur mi a story more thanonce, either through trebbling, or additional moves, roughly reflectingthe tal.e's length. On the other hand, the number of different types offunctions suggests something of the tale's richness while number of movesmay indicate complexity. As can be seen from Table 3, Salt and Squawk tothe Moon, Little Goose are equally rich, though Salt is shorter andsom bat more complex. They differ considerably, however, from MagicPo id e 2ot, a fairly straightforward and brief story with a slight
onic twist in the second move. Both Salt aad Squawk to the Moon;.Little Goose contain parallel action and multiple embedding. WhileSquawk to the Moon, Little Goose embodies the simple, but clear, moralambiance of a fable for children, Salt has all the atmospherics of a trueRussian fairy tale. Thus, each story constituted a rather differentexperience for each retelling.
Table 3
Number of Functions, Types, and Moves in Three Stories
Stories Functions Types Moves
Squawk to the Moon, Little Goose 29 18 2
Magic Porridge Pot 15 12 2
Salt 22 18 3
1 8 a5
-152-
Analysis of Story Structure Data
Both multivariate and univariate analyses of variance were employed
for story structure comparisons. For the multivariate analyses, as withcohesion, computet program CANOVA (Clyde Computing Services, 1973) was
used. This program tests for significant differences with Wilks'sliklihood ratio transformed to Rao's approximate F. Significantmultivariate differences were followed-up w".th univariate analyses ofvariance.
Number of functions, function types, and moves, served as dependentvariables in six complementary multivariate analyses of varianceperformed on the story structure data. in the first of these analyses,144 score: for each dependent variable were organized into a mixeddesign, where sex (six males and six females) and dialect (six vernacularand six nonvernacular) served as between-subjects comparisons, and wheremodes of discourse (retelling and dictation) and observation periods(Spring 1979, Autumn 1979, Spring 1980) constituted the within-subjectscomparisons. This study was designed to compare factors within the urbanschool setting. Similar design arrangements were employed in a secondanalysis whose purpose was to compare the urban with the suburban schoolcontrolling for dialect. While only middle class children from the twoschools were compared, the two populations did differ on the index ofstatus characteristics with t (24 df) = 2,79 (.2 < .01). Children fromthe suburban sLnool averaged from middle to upper-middle class on the"index" (M = 33.33; SD = 4.37). While those from the urban school,averaged somewhat higher scores on the scale (M = 38.33; SD = 7.79). Thetwo populations had been equated on the scale at the outset of the study,but because of subject mortality and replacement, this initial equalitywas lost necessitating a school comparison. For this comparirandependent variables were organized into a 2 x 2 x 2 x 3 mixed designwhere sex and school were the between-subject.; factors and where modesand observations were the within-subjects factors. A third multivariateanalysis of variant: then was employed to examine'only the suburbanschool. As before, number of functions, function types, and moves, wereorganized into a mixed design with one between-subjects comparison--sex(six males and six females)--and two within-subjects compariscns"modesand observations.
Three additional multivariate analyses of variance focused upondictation. Retetting was removed as a comparison in order to obtain aclearer view of dictation over the three observation periods--retellingdifferenc. s having potentially spurious origins in the 'varianceassociated with apriori story differences. In all other respects, designgoals and arrangements were identical to those reported above.
.Significant MANOVA test statistics were followed up by univariateanalyses of variance. These designs compared the same variables,organized in the same ways reported above, for the multivariate analyses.Significant univariate F ratios were subjected to Geiseer-Greenhouse
-153-
182
conservative F corrections for repeated-measures designs. Post hoc
comparisons were made using Tukey's H.S.D. procedure.
Procedures for Coding Concept of Message
Two additional unlvariate analyses of variance were performed onfunctions and function types from texts produced by a sample of subjectswho were able to compose unequivocal fictional narratives. Just 14subjects were able to do so by mid-first grade. This number rose to 27at the end of grade two. The point of these two analyses was to obtaindevelopmental data controlled rigorously for genre. Other genres of textwere excluded from these analyses to eliminate genre as a contaminatingsource of variance.
During the early stages of becoming literate, young children beginto gain control over basic concepts about the organization of surfacefeatures of written language. They learn the specifics of how textsconvey information, e.g., that the groups of letters, not the pictures,carry the message, or that particular patterns of letters correspond toparticular spoken words (Clay, 1975; Henderson, 1980). Simultaneously,they also internalize and use the rules governing direct physical aspectsof text, e.g., conventions of spacing and directionality. As part ofthis study, samples of children's writings were examined to see howch' 'ten differed in their understanding and use of these principles.
Sets of exhaustive, mutually exclusive categories, were developedfor each of the three dimensions of Concept of Message, Directionality,and Spacing. (These will be described in greater detail in the section,Results and Discussion: Conventions of Print.) Based on their writingsamples, each subject was classified as being in one category, for eachdimension, for each of the five observations. Because of the explicitnature of the categories (e.g., percent of word boundaries observed,string of random letters), a single investigaZion--working with thewriting samples and data collectors' written comments--classified thedata. No assumptions have been made about the linear or progressivenature of the categories. It was expected however, that, in a generalway, subjects would be classified in the higher number categories as theygained more control over the conventions. The number and percentage ofchildren per category was tabulated by sex, dialect, school, andobservation. These data will be reported\in Chapter 3.
183-154-
APPENDIX B
Cohesion MANOVAs by Mode and Gradefor Urban and Suburban Schools
The tables which follow are included in this report for the reader
who may wish to examine the bases for decisions to analyze dictation,
writing, and grade level data in separate multivariate analyses. These
tables show significant, multivariate, urban-school effects for grade,
dialect, mode and observation; significant second-order interactions
for grade by sex by thseriation, grade by mode by observation, and
sex by mode by observation; and significant first-order interactions
for sex by mode, sex by observation and mode by observation.
Tables containing relevant means and standard deviations also are
included in Appendix B.
Table 8.1Cohesion MANOVA by Grade, Dialect, Sex, Mode, and Observation in
Dictation and Writing at Urban School (K-1/1-2)
Source df dfIRP dfERR E.<
Between Subjects 39
Grade (A) 1 5.00 28.00 4.14 .006
Dialect (B) 1 5.00 28.00 3.01 .03
Sex (C) 1 5.00 28.00 .55 .74
Grade by Dialect (AB) 1 5.00 28.00 2.29 .07
Grade by Sex (AC) 1 5.00 28.00 3.90 .008
Dialect by Sex (BC) 1 5.00 28.00 1.11 .38
grade X Dialect X Sex 1 5.00 28.00 .70 .63
(ABC)
S/ABC 32
Within Subjects 20C
Mode (D) 1 5.00 28.00 27.75 .001
Grade X Mode (AD) 5.00 28.00 2.86 .03
Dialect X Mode (BD) 1 5.00 28.00 2.79 .04
Sex X Mode (CD) 1 5.00 28.00 1.30 .29
Grade X Dialect XMode (ABD) 1 5.00 28.00 4.02 .007
Grade X Sex X Mode (ACD) 1 5.00 28.00 .30 .91
Dialect X Sex X Mode (BCD) 1 5.00 28.00 .77 .58
Grade X Dialect X SexX Mode (ABCD) 1 5.00 28.00 1.16 .36
SD/ABC 32
Observation (E) 2 10.00 120.00 7.14 .001
4.00 60.50 4.12 .005
Grade X Observation (AE) 2 10.00 120.00 2.10 .03
4.00 60.50 .96 .44
Dialect X Observation (BE) 2 10.00 120.00 .60 .81
4.00 60.50 .24 .91
Sex X Observation (CE) 2 10.00 120.00 1.55 .13
4.00 60.50 1.57 .20
Grade X Dialect XObservation (ABE) 2 10.00 120.00 1.04 .41
4.00 60.50 .43 .79
Cohesion MANOVA at Urban School (Cont.)
Source df dflUT dfERR F E.<
Grade X Sex XObservation (ACE) 2 10.00 120.00 .84 .59
4.00 60.50 .40 .81
Dialezt X Sex XObservation (BCE) 2 10.00 120.00 1.46 .18
4.00 60.50 .83 .51
Grade X Dialect X SexX Observation (ABCE) 2 10.00 120.00 .13 .99
4.00 60.50 .05 .99
SE/ABC 64
Mode X Observation (DE) 2 10.00 120.00 4.05 .001
trade X Mode X
4.00 60.50 2.30 .07
Observation (ADE) 2 10.00 120.00 1.46 .16
4.00 60.50 .40 .81
Dialect X Mode XObservation (BDE) 2 10.00 120.0( .55 .85
4.U0 60.50 .03 .99
Sex X Mode X
Observation (CDE) 2 10,00 120.00 .72 .70
4.00 60.50 .43 .79
Grade X Dialect X Mode XObservation (ABDE) 2 ,) 10.00 120.00 .81 .60
4.00 60.50 .40 .81
Grade X Sex X Mode XObservation (ACDE) 2 10.00 120.00 .47 .90
4.00 60.50 .28 .89
Dialect X Sex X Mode XObservation (BCDE) 2 10.00 ' 120.00 .83 .61
4.00 60.50 .40 .81
Grade X Dialect X Sex X 2 10.00 120.00 .50 .89Mode X Observation (ABCDE) 4.00 60.50 .39 .82
SDE/ABC 64,
TOTAL 239
187-158-
Table B.2
Means and Standard Deviations for Cohesion MAMA (Transformed Variables)for Urban School (K-1/1-2) -- by Grade, Dialect, and Mode
Grade Dialect Mode R Exo Ref Reference gilipsis Conjunction Lexical
K-1
Vernacular
Nonvernacular1
,....
Ln.41,
1-2
a
183
Vernacular
Nonvernacular
Dialect Means:
Mode Means:
M SD M SD M SD M
.06 .18 .38 .30 .01 .02 .10
.08 .24 .38 .36 10.01 .02 .10
Dictation .05 .05 .42 '.12 .01 .02 .18
Writing .11 .33 .34 .50 .00 .01 .02
1..051
.09 .38 .22 .01 .03 .10.
Dictation .05 .10 .42 .15 .02 .03 .16
Writing .04 .09 .34 .27 .01 .02 .05
.08 .26 .36 .20 .02 .04 .15
.10 .24 .33 .22 .01 .03 .13
Dictation .05 .08 .42 .16 .02 .04 .19
Writing .15 .32 .23 .23 .01 .01 .07
.06 .21 .40 .18 .03 .06 .17
'Dictation .02 .02 .40 .07 .02 .02 .21
Writing .10 .30 .41 .25 .04 .07 .13
Vernacular .09 .24 .35 .30 .01 .02 .11
Nonvernacular .05 .17 .39 .20 .02 .04 .14
Dictation .04 .07 .41 .13 .02 .03 .18
Writing .10 .28 .33 .33 .01 .04 .07
SD M SD
.11 .29 .23
.11 .25 .23
.08 .37 .20
.06 .12 .20
.12 .33 .213
.12 .38 .13
.09 .27 .26
.11 .32 .17
.13 .31 .19
.10 .34 .09
.12 .28 .25
.09 .32 .15
.05 .38 .08
.10 .27 .18
.12 .28 .22
.11 .32 .181
.09 .37 .13
.10 .24 .23
MU 1101 all MR las Ili me MI sr MI my ink Eli is ire *a
am in fam so tug me in En MI MO all OM as 1.1 3111 (111.1
Table B.3
Means and Standard Deviations of Cohesion MANOVA (Transformed Variables)in Dictation and Writing for Urban School (*-1/1-2) -- by Grade and Observation
Grade Observation R Rao Ref Reference Ellipsis Conjunction Lexical..
M SD M SD M SD M SD M SD
K-I I .09 .13 .32 .26 .01 .03 .08 .10 .20 .18
2 .02 .03 .38 .30 .02 .02 .14 .14 .24 .19
3 .08 .28 .44 .33 .00 .01 .09 .09 .42 .25
1-2 I .18 .36 .30 .21 .01 .05 .12 .10 .27 .19
2 .02 .04 .40 .22 .03 .04 .15 .11 .31 .18
3 .03 .08 .41 .17 .02 .05 .18 .11 .37 .13tt-a.0i Observation
Means: I .13 .28 .31 .24 .01 .04 .10 .10 .24 .19
2 .02 .04 .39 .26 .02 .03 .14 .12 .28 .18
3 .06 .20 .42 .26 .01 .04 .14 . .11 .39 .20
190 19i
Table 8.4
Means and Standard Deviations of Cohesion MANOVA (TransformedVariables) for Urban School (K4/1-2) -- by Grade and Mode
i
,-ch-I
Grade Mode R Exo Ref Reference tllipsis Conjunction Lexical
K-1
1-2
Grade Means:
DictationWriting
DictationWriting
K-1
1-2
M
.06
.05
.07
.08
.04
.12
.06
.08
SD
.18
.08
.24
.23
.06
.31
.18
.26
M
.38
.42
.34
.36
.41
.32
.38
.36
SD
.30
.13
.40
.20
.12
.26
.30
.20
M
.01
.02
.01
.02
.02
.02
.01
.02
SD
.02
.03
.02
.04
.03
.06
.02
.04
M
.10
.17
.03
.15
.20
.10
.10
.15
SD
.11
.10
.08
.13
.08
.11
.11
.11
M
.29
.38
.20
.32
.36
.27
.29
.32
SD
.23
.17
.24
.17
.09
.22
.23
.17
193
IS tie ifs We IMe aft my lab Mu or ow an
- ow Mt SIM 111111 lilt Mt Ili la 1111111 MI Int Mir
Table B.5
Means and Standard Deviations of Coheaion MANOVA (Transformed Variables)for Urban School (*-1/1-2) -- by Dialect and Mode
Dialect Mode R Exo Ref Reference Ellipsis Conjunction Lexical
M SD ii a it SD ii SD M ',SD
Vernacular Dictation .05 .06 .42 .14 .02 .03 .18 .09 .36 .15
Writing .13 .32 .29 .39 .00 .01 .04 .10 .20 .24
Nonvernacular.Dietation .04 .07 .41 .11 .02 .03 .18 .09 .38 .11
...
at Writing .07 .22 .37 .26 .03 .06 .09 .11 .27 .2214i
194195
Table 11,6
Hanna and Standard Deviations of Coltaalon MAMMA (Transformed VariableaJ
Grnd,
K-1
Sex
rot Urban School (R-1/1-7% --
R Exo Ref Reference
SD N SD
by Grade end SK
Ellipaie
K SD
Conjunction
H SD
Lexical
!"1SU
MatOA .06 .14 .35 .28 .01 .03 .10 .12 .31 .25
FOinalue .07 .21 .41 .31 .01 .02 .10 .11 .26 .19
1-2
Malzd .10 .30 .36 .22 .01 .04 .17 .12 .29 .17
Females .05 .12 .37 .19 .03 .05 .13 .11 .35 .17
19?
196'
Mil UN all MI MO allf Ile 1111 IWO INN @MI MI AO MB INN OS MI IMO
lea ala 0011 Mal WO IS MI ems sae ow En as um es re um es En ma
Table 11.7
Menne and standard Deviation° of 'Cohesion MANOVA (Transformed Variables)for Urban School (K-1/1-2) -- by Mode and Observation
i
heCAt.1
Mode Observation R Exo Ref Reference Ellipsis Conjunction Lexical
Dictation1
1
3
Writing1
2
3
SD
.08 .10
.03 .03
.02 .03
.19 .37
.01 .04
.09 28
M
.39
.43
.42
24.34
.41
SD
.12
12.15
.29
.35
.34
SD
.01 .03
.03 .03
01 .02
.01 .05
.01 .03
.02 .05
M
.17
.21
.17
.02
.08
10
SD
.08
.11
.08
.06
.10
.12
M
.36
.33
.42
.11
.23
.37
SD
.10
.07
.19
.17
.24
.21
193
193
Table B.8
Cohesion MANOVA by Grade, Sex, Mode, and Observationin Dictation and Writing at Suburban School (K-1/1-2)
Source df
Between Subjects 23
Grade (A) 1
Sex (B) 1
Grade X Sex (AB) 1
S/AB 20
Within Subjects 120
Mode (C) 1
Grade X Mode (AC) 1.
Sex X bode (BC) 1
Grade X Sex X Mode (ABC) 1.
SC/AB 20
Observation (D) 2
Grade X Observation (AD) 2
Sex X Observation (BD) 2
Grade X Sex X 2
Observation (AB D)
SD/ AB 40
Mode X Observation (CD) 2
Grade X Mode X 2
Observation ('taD)
Sex X Mode X 2
Observation (BCD)
Grade X Sex X Mode X 2Observation (ABCD)
SCD/AB 40
TOTAL 143
dfHTP dfERR <
5.00 16.00 2.53 .07
5.00 16.00 2.53 .0/
5.00 16.00 1.20 .13
5.00 16.00 5.77 .003
5.00 1.6.00 1.76 .1.8
5.00 1.6.00 3.20 .03
5.00 16.00 1.38 .28
10.00 72.00 4.28 .001.
4.00 36.50 1.56 .21
10.00 72.00 3.04 .003
4.00 36.50 .90 .47
10.00 72.00 2.14 .03
4.00 36.50 .55 .70
10.00 72.00 2.01 .04
4.00 36.50 .98 .43
10.00 72.00 3.38 .001
4.00 36.50 .81 .53
10.00 72.00 2.56 .01
4.00 36.50 .48 .75
10.00 72.00 2.38 .02
4.00 36.50 1.04 .40
10.00 72.00 1.61 .12
4.00 36.50 .97 .44
20O
-165-
11
ell all Ili 1111 MI WI WM MI go MS 1111111 MI Mt NS NW MI MI 1111I
Table 8.9
Means and Standard Deviations of Cohesion MANNA (Transformed Variables)in Dictation and Writing for Suburban School -- by Grade and Observation
Grade Observation
K-1 1
2
3
1-2 1
2
3
Grade Means:K-1
1-2
Observation Means:
1
2
3
R Exo Ref Reference Ellipsis Conjunction Lexical
MP.
M SD M SD M SD M SD M SD
.03 .07 .46 .34 .01 .02 .12 .10 .26 .19
.04 .15 .41 .20 .08 .32 .15 .13 .38 .16
.00 .01 .47 .14 .03 .06 .15 .09 .38 .12
.03 .07 .52 .27 .01 .01 .14 .11 .37 .18
.02 .03 .40 .10 .02 .03 .21 .08 .37 .10
.01 .01 .38 .08 .01 .02 .17 .07 .45 .10
.02 .09 .45 .24 .04 .19 .14 . .11 .34 .17
.02 .05 .44 .18 .01 .02 .17 .09 .39 .14
.03 .07 .49 .31 .01 .02 .13 .10 .32 .19
.03 .11 .41 .16 .05 .23 .18 .11 .37 .13
.00 .01 .43 .43 .02 .04 .16 .08 .41 .11
20i202
Table B.I0
Means and Standard Deviations for Cohesion MANOVA (Transformed Variables) in Dictationand Writing for Suburban School -- by Grade, Sex, and Observation
Grade Sex Observation R Exo Ref Reference Ellipsis Conjunction Lexical
M SD M SD M SD M SD M SD
K-I Males I .04 .10 .60 .37 .01 .03 .13 .08 .29 .21
2 .01 .02 .36 .19 .15 .45 .14 .12 .41 .17
3 .00 .00 .47 .13 .02 .04 .17 .09 .37 .11
Females 1 .01 .03 .32 .26 .00 .01 .11 .11 .23 .18
2 .06 .21 .46 .20 .01 .01 .16 .14 .35 .15
3 .00 .01 .47 .15 .03 .08 .14 .10 .38 .13
*.s
a.-4
1-2 Males I
2
.05
.02
.10
.03
.58
.42
.36
.09
.00
.01
.01
.02
.15
.24
.12
.07
.36
.33
.22
.091 3 .01 .01 .37 .09 .01 .01 .21 .07 .43 .09
Females 1 .01 .02 .46 .13 .01 .02 .13 .10 .43 .09
2 .01 .02 .38 .11 .04 .03 .18 .09 .41 .11
3 .01 .01 .39 .06 .02 .02 .14 .05 .46 .12
Observation Means: I .03 .07 .49 .31 .01 .02 .13 .10 .32 .19
2 .03 .11 .41 .16 .05 .23 .18 .11 .37 .13
3 .00 .01 .43 .12 .02 .04 .16 .08 .41 .11
203 204
ON IND OM INS MI IMP Ile INN III MB MI NIP Mil MO 111 MI Oil MIR AM
APPENDIX C
Cohesion MANOVAs for School, Mode and Observation (K-1)
The decision to analyze kindergarten cohesion data by individual
schools was based upon a series of school comparisons. In these multi-
variate analyses of variance and companion discriminant-univariate
follow-ups, cohesion data were organized into a design where school and
(sex were treated as betweensubjects factors and mode and observation
were treated as within-subjects factors. As shown in Table C.l,
significant MANOVA test statistics were indicated for the school factor
and two school-related, first-order interactions: school by observation
and school by mode. Follow-up multivariate analyses for both dictation
and writing produced no significant effect for dictation (See Table C.2 )
but a significant second-order interaction in writing for school by sex
by observation (See Table C.3 ). Univariate analyses of variance indicated
a significant second-order interaction for reference: F (2, 32) = 5.68,
E.< .008. This interaction is graphed in Figure C.l. Tukeyvagl lass,
tests indicated that suburban boys employed higher proportions of reference
ties at observation one than suburban girls, urban girls, or urban boys.
First-order, school by observation and sex by observation interactions were
not significant.
The significant multivariate test statistic for the observation
factor in writing (See Table C.4 ) was followed-up by the usual discrimi-
nant and univariate analyses (See Table C.5 ). Discriminant analysis
indicated that lexical cohesion made the largest contribution to the
206-169-
function followed by reference and ellipsis--the latter two, failing
to achieve significance on univariate tests, probably contributing to
the discrimination through their contrasting relative stability. Re-
stricted exophoric reference, conjunction, and lexical cohesion did
achieve designated univariate alpha levels (See Table C.5.) for the
observation factor. Exophoric reference proportions decreased sig-
nificantly over observations--the sharpest drop at observation two the
beginning of second grade. Conversely, conjunction proportions increased
at the beginning of second grade, then dropped slightly. Lexical cohesion
proportions increased uniformly over observations--the significant
difference shown between grade one and the end of grade two. Means and
standard deviation for the observation factor in writing are displayed
in Table C.6.
The cohesion MAMA for the urban school population produced
significant multivariate test statistics for the mode and observation
factors (both roots) and for the firsL order mode by observation inter-
action. Therefore, separate follow-up MANOVAs were perCormed for each
mode, that is dictation and writing (See Table C.? ).
The MANOVA for dictation at the urban school indicated a significant
multivariate test statistic for both roots of the observation factor (See
Table C.8 ), and no first- or second-order interaction effects. Univari-
ate and discriminant follow-ups (See Table C.9 ) indicated that lexical
cohesion and reference contributed strongly to the function through the
first roottand all variables except ellipsis contributed strongly to the
function through the second root. Reference was the only univariate
variable not to achieve significance for the observation factor in dictation.
Means and standard deviations, given in Table C.10, indicate a significant
decrease in exophoric reference proportions at observation two and a
significant decrease in ellipsis at observation three. Conjunction
proportions decreased at observation three while lexical cohesion pro-
portions increased significantly at observation three.
The urban school MANOVA for writing produced a significant multi-
variate test statistic for the dialect and observation factors (See
Table C.II ). The usual follow-up procedures indicated lexical cohesion
and reference proportions, in that order, made the strongest contribution
to the discrimination for the dialect factor; however, only lexical
cohesion reached a significant univariate alpha criterion (See Table C.I2 ).
Tukeyaliat bnc,comparisons of mean differences fet kindergarten lexical
cohesion proportions (See Table C.I4 for means and standard deviations)
indicated that nonvernacular subjects used higher proportions of lexical
cohesion in their texts than vernacular subjects. For the observation
factor, lexical cohesion and conjunction proportions achieved significant,
univariate, alpha levels (See Table C.13)--increases in lexical proportions
having occurred at observation three; and significant increases in con-
junction proportions having occurred at observation two, persisting through
observation three (See Table C.10,
Similarly, suburban-school cohesion data for dictation and writing
were first compared in a multivariate design (MANOVA) where sex was the
between-sub jetAs factor and where mode and observation were within-subjects
factois. Significant MANOVA test statistics were obtained for the mode
and observation factor as well as the first-order mode by observation
interaction (See Table C.15 ). Significant mode and mode by observation
effects led to the decision to analyze dictation and writing in separate
MANOVAs. The MANOVA for dictation data produced no significant main
or interaction effects (See Table C.16 ) for kindergarten cohesion
proportions at the suburban school. However, the suburban-school
cohesion MANOVA for writing produced a significant test statistic for
the observation factor (See Table C.17). Discriminant and univariate
follow-up analyses indicated that lexical cohesion proportions made the
largest contribution to the discrimination, followed by ellipsis,
exophoric reference and conjunction (See Table C.18). Only lexical
cohesion and conjunction, however, were implicated in the univariate
analyses of variance. Mean differences (See Table C.19) demonstrated
conjunction proportions increased significantly at observation two and
continued to climb gradually at observation three. The same pattern was
indicated for lexical cohesion proportions. Tukey zat bac.comparisons
indicated that lexical cohesion proportions increased significantly at
observation two followed by a smaller increase at observation three--
both, however, significantly higher than observation one.
Table C.1
Cohesion MANOVA in Dictation and Writing by School,Sex, Mode, and Observation -- (K-1)
Source df dfHYP dfERR F E.
Between Subjects 19
School (A) 1 5.00 12.00 5.82 .006
Sex (B) 1 5.00 12.00 1.37 .30
School X Sex (AB) 1 5.00 12.00 .73 .62
S/AB 16
Within Subjects 100
Mode (C) 1 5.00 12.00 13.05 .001
School X Mode (AC) 1 5.00 12.00 3.39 .04
Sex X Mode (BC) 1 5.00 12.00 .79 .58
School X Sex X Mode (ABC) 1 5.00 12.00 .29 .91
SC/AB 16
Observation (D) 2 10.00 56.00 4.81 .001
4.00 28.50 2.09 .11
School X Observation (AD) 2 10.00 56.00 2.14 '.04
4.00 28.50 .90 .48
Sex X Observation (BD) 2 10.00 56.00 1.09 .39
4.00 28.50 .68 .61
School X Sex X Observation (ABD) 2 10.00 56.00 .70 .72
4.00 28.50 .62 .65
SD/AB 32
Mode X Observation (CD) 2 10.00 56.00 3.02 .004
4.00 28.50 1.42 .25
School. X Mode X Observation (ACD) 2 10.00 56.00 1.31 .25
4.00 28.50 .63 .65
Sex X Mode X Observation (BCD) 2 10.00 56.00 1.00 .45
4.00 28.50 .51 .73
School X Sex X Mode X 2 10.00 56.00 .50 .88
Observation (ABCD) 4.00 28.50 .33 .86
SCD/AB 32
_TOTAL 3.3.9
-173-1u
Table C.2
Cohesion MANOVA in Dictation by School, Sex, andObservation -- (K-1)
Source df dfHYP dfERR E.(
Between Subjects 19
School (A) 1 5.00 12.00 2.16 .13
Sex (B) 1 5.00 12.00 1.18 .37
School X Sex (AB) 1 5.00 12.00 .51 .76
S /AS 16
Within Subiects 40
Observation (C) 2 10.00 56.00 1.70 .11
4.00 28.50 1.74 .17
School It Observation (AC) 2 10.00 56.00 1.89 .07
4.00 28.50 1.31 .29
Sex X Observation (BC) 2 10.00 56.00 .34 .97
4.00 28.50 .20 .94
School X Sex X Observation 2 10.00 56.00 .72 .70(ABC) 4.00 28.50 .53 .72
SC/AB 32
TOTAL 59
211
-174-
1
Table C.3
Cohesion MANOVA in Writing by School, Sex, andObservation -- (K-1)
Source df dfHYP dfERR F
Between Subjects 19
School (A) 1 5.00 12.00 1.99 .15
Sex (B) 1 5.00 12.00 .31 .90
School X Sex (AB) 1 5.00 12.00 1.05 .43
S/AB 16
Within Subjects 40
Observation (C) 2 10.00 56.00 4.18 .001
4.00 28.50 1.08 .38
School X Observation (AC) 2 10.00 56.00 1.53 .15
4.00 28.50 .57 .68
Sex X Observation (BC) 2 10.00 56.00 .89 .55
4.00 28.50 .53 .71
School X Sex X 2 10.00 56.00 2.72 .008
Observation (ABC) 4.00 28.50 1.89 .14
SC/AB 32
TOTAL 59
212
90-60-70-60-50-40-30-20-0-1
0-
03
02
01
REFERENCE
90-60-70-60-
01 50-02 40-
30-03 20-
10-
Male Female
URBAN SCHOOL
90-60-70-60-50-40-30-20-10-0-
01
03
02
0-
Q1
03
02
REFERENCE
Male Female
SUBURBAN SCHOOL
REFERENCE
2
03
01
Male Female
SCHOOLS
Figure C.1. Observation as a Function of Sexand School for School by Sex byObservation Interaction in Retelling (K-1)
-176- 213
1
1
Table C.4
Standard Discriminant Function Coefficients and Univariate ANOVAson Use of Cohesion Categories in Writing for School by Sex by
Observation -- (K-1)
Cohesion Category
DiscriminantFunction
CoefficientsUnivariate F Tests(2, 32) 2:4
Restricted Exophoric Reference .469 1.72 .20
Reference 1.001 5.68 .008
Ellipsis .570 1.43 .25
Conjunction .303 1.10 .35
Lexical Cohesion .659 1.19 .32
Table C.5
Standard Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Writing for Observation -- (K-1)
Cohesion Category
DiscriminantFunction
Coefficients
Univariate F Tests(2, 31) E. 4
Restricted Exophoric Reference .192 6.53 .004
Reference -.528 .15 .86
Ellipsis -.572 1.03 .37
Conjunction -.467 4.54 .02
Lexical Cohesion -.821 18.78 .001
Table C. 6
Means and Standard Deviations of Cohesion Categories in Writingby School, Sex, and Observation .-
School Sex 'Observation R Exo Ref Reference Ellipsis Conjunction Lexical
Urban
Suburban
Males
Females
Males
Females
Observation Means:
i SD I SD if SD Y SD X SD
.05 .10 .38 .22 .01 .03 .10 .12 .33 .21
.03 .09 .33 .25 .01 .03 .11 .14 .33 .24
1 .05 .11 .17 .38 .00 .00 .00 .00 .00 .00
2 .00 .00 .23 .27 .00 .00 .02 .04 .19 .26
3 .00 .00 .46 .23 .01 .02 .09 .13 .49 .11
.06 .10 .43 .18 .02 .02 .10 .10 .33 .19
1 .15 .16 .49 .20 .00 .00 .00 .00 .22 .25
2 .03 .04 .41 .24 .03 .04 .13 .15 .22 .15
3 .00 .00 .29 .17 .02 .02 .03 .04 .50 .30
.01 .05 .46 .25 .04 .20 .15 .11 .34 .26
.02 .06 .49 .29 .07 .29 .15 .11 .34 .17
1 .08 .15 .81 .43 .02 .03 .04 .04 .17 .17
2 .01 .02 .32 .30 .33 .69 .13 .17 .36 .25
3 .00 .00 .46 .21 .00 .00 .19 .12 .39 .10
.01 .02 .44 .22 .01 .05 .14 .11 .33 .15
1 .02 .04 .26 .35 .00 .00 .02 .05 .12 .17
2 .00 .00 .59 .26 .00 .00 .06 .09 .41 .12163 .00 .00 .52 .24 .05 .11 .10 .10 .37 .17
1 .07 .12 .43 .31 .01 .03 .10 .10 .24 .19
-2 .01 .02 .39 .21 .06 .25 .16 .14 .33 .16
3 .01 .02 .44 .18 .02 .05 .13 .10 .42 .17
Table C.7
Cohesion HANOVA in Dictation and Writing by Dialect
Sex, Mode, and Observation -- at Urban School (1(-1)
Source df dfHYP dfERR P 2:4
Between Subjects 19
Dialect (A) 1 5..00 12.00 1.67 .22
Sex CS) 1 5.00 * 12.00 1.48 .27
Dialect X Sex (AB) 1 5.00 12.00 .58 .71
5 /AB 16
Within SubJectt 100
Mode (C) 1 5.00 12.00 19.25 .001
Dialect X Made (AC) 1 5.00 12.00 1.77 .19
Sex X Mode (BC) 1 5.00 12.00 .88 .53
Dialect X Sex X Mode (ABC) 1 5.00 12.00 1.32 .32
=AB 16
Observation (D) 2 10.00 56.00 6.79 .001
4.00 28.50 4.13 .009
Dialect X Observation (AD) 2 10.00 56.00 1.09 .39
4.00 28.50 .61 .66
Sex X Observation (BD) 2 10.00 56.00 1.42 .20
4.00 28.50. 1.74 .17
Dialect X Sex X 2 10.00 56.00 .98 .47
Observation (ABD) 4.00 28.50 .67 .62
SD/AB 32
Mode Z Observation (CD) 2 10.00 56.00 2.45 .02
4.00 28.50 .25 .91
Dialect X Made X 2 10.00 56.00 1.10 .38
Observation (ACD) 4.00 28.50 .46 .76
Sex X Mode X Observation 2 10.00 56.00 .57 .83
(BCD) 4.00 28.50 .22 .93
Dialect X Sex X Mode X 2 10.00 56.00 .99 .46
Observation (ABCD) 4.00 28.50 .27 .90
SCD/AB 32
TOTAL 119
217
Table C.8
Cohesion MANOVA in Dictation by Dialect, Sex, andObservation -- at the Urban School (E-1)
Source df dfHYP dfERR 2.4
Between Subjects 19
Dialect (A) 1 5.00 12.00 .14 .98
Sex (B) 1 5.00 12.00 1.11 .41
Dialect X Sex (AB) 1 5.00 12.00 .65 .67
SLAB 16
Within Subjects 40
Observation (C) 2 10.00 56.00 3.36 .002
4.00 28.50 3.39 .02
Dialect X Observation (AC) 2 10.00 56.00 .70 .72
4.00 28.50 .37 .83
Sex X Observation (BC) 2 10.00 56.00 .73 .69
4.00 28.50 .40 .81
Dialect X Sex X Observation 2 10.00 56.00 .68 .74
(ABC) 4.00 28.50 .37 .83
SC/AB 32
TOTAL 59
Table C.9
Standard Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Dictation for Observation at Urban
School (K-1)
Discriminant FunctionCoefficients Univariate F Tests
Cohesion Category Root Root (2, 32) E.<1 2
Restricted Exophoric Reference 0.034 -1.337 5.24 .01
Reference -1.332 -3.366 .12 .88
Ellipsis 0.326 -0.736 3.05 .06
Conjunction -0.649 -2.913 3.84 .03
Lexical Cohesion -1.664 -2.971 5.57 .008
219-181-
Table C,10
Mans and Standard Deviations of Cohesion Categoriesby Mode and Observation at the Urban School (1(1)
Mode Observation R Exo Ref Reference Ellipsis Conjunction Lexical
lr SD ir SD lr SD lir SD X SD
Dictation
Writing
ObservationOverall:
.05 .08 .42 .13 .02 .03 .17 .10 .3$ .17
1 .10 .12 .42 .11 %02 .04 .15 .09 .32 .09
2 .03 .04 .42 .14 .02 .02 .21 .12 .33 .07
3 .03 .03 .40 .18 .01 .01 .14. .08 .47 .24
.07 .24 .34 .40 .01 .02 .03 .08 .20 .24
1 .08 .15 .23 .33 .00 .00 .00 .00 .07 .16
2 .01 .02 .33 .40 .01 .03 .06 .10 .15 .22
3 .14 .39 .47 .44 .01 .01 .04 .08 .36 .25
1 .09 .13 .32 .26 .01 .03 .08 .10 .20 .18
2 .02 .03 .38 .30 .02 .02 .14 .14 .24 .19
3 .08 .28 .44 .33 .00 .01 .09 .09 .42 .25
22 t)
Li 0 Al MI MI SO la MN NO lin OM la NIS 1111111
Table C.11
Cohesion MANOVA by Dialect, Sex, and Observationin Writing for Urban School (K-1)
Source df dfHYP dfERR P.4
Between Subjects 19
Dialect (A) 1 5.00 12.00 3.53 .03
Sex (B) 1 5.00 12.00 0.76 .60Dialect X Sex (AB) 1 5.00 12.00 1.49 .26
S/AB 16
Within Subjects 40
Observation (C) 2 10.00 56.00 5.02 .001
4.00 28.50 2.22 .09
Dialect X Observation 2 10.00 56.00 1.23 .29
(AC) 4.00 28.50 .66 .62
Sex X Observation (BC) 2 10.00 56.00 1.16 .34
4.00 28.50 .94 .46
Dialect X Sex X 2 10.00 56.00 1.37 .22
Observation (ABC) 4.00 28.50 .87 .49
.SC/AB 32
TOTAL 59
Table C.I2
Standard Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Writing for Dialect at Urban School (1C1)
Cohesion Category
DiscriminantFunction
CoefficientsUnivariate F Tests(I, 16) 2.4
Restricted Exophoric Reference 0.607 1.40 .26
Reference 1.215 .00 .99
Ellipsis -0.410 3.57 .08
Conjunction 0.069 1.55 .23
Lexical Cohesion -1.301 6.95 .02
Table C.I3
Standard Discriminant Function Coefficients and Univariate ANOVAs onUse of Cohesion Categories in Writing for Observation at Urban School (K -I}
Cohesion Category
DiscriminantFunction
CoefficientsUnivariate F Tests(2, 32)
Restricted Exophoric Reference 0.411 1.39 .26
Reference 0.747 2.66 .09
Ellipsis -0.042 2.37 .11
Conjunction 0.067 3.19 .05
Lexical Cohesion 1.024 13.10 .001
222
am as es so me um us um um or on am
Table C.14
Means and Standard Deviations of Cohesion Categories in
Writing by Dialect at Urban School (K-I)
Dialect It Exo Ref Reference Ellipsis Conjunction Lexical
SD X SD K SD X SD I SD
Vernacular .08 .24 .38 .36 .01 .02 .10 .11 .25 .23
onveraacular .05 .10 .38 .22 .01 .03 .10 .12 .33 .21
Table C.15
Cohesion MANOVA in Dictation and Writing bySex, Mode, and Observation -- Suburban School (K-1)
Source df flan dfERR R.<
Between Subjects 13
Sex (A) 1 5.00 8.00 .66 .67
S/A 12
WithiajubJects 70
Mode (B) 1 5.00 8.00 4.04 .04
Sex X Mode (AB) 1 5.00 8.00 1.29 .36
SB/A 12
Observation (C) 2 10.00 40.00 3.34 .003
4.00 20.50 .73 .58
Sex X Observation (AC) 2 10.00 40.00 .74 .69
4.00 20.50 .45 .77
SC/A 24
Mode X Observation (BC) 2 10.00 40.00 2.48 .02
4.00 20.50 .66 .63
Sex X Mode X 2 10.00 40.00 .97 .49
Observation (ABC) 4.00 20.50 .63 .64
SBC/A 24
TOTAL 83
Oir,....
Table C.16
Cohesion MANOVA in Dictation by Sex andObservation at the Suburban School (X,.1)
Source df dfHYP dfERR P E.
Between Subjects 13
Sex (A) 1 5.00 8.00 1.42 .31
S/A 12
Within Subjects 28
Observation (B) 2 10.00 40.00 1.46 .19
4.00 20.50 .43 .79
Sex X Observation (AB) 2 10.00 40.00 .65 .76
4.00 20.50 .35 .84
SB /A 24
TOTAL 41
Table C.17
Cohesion MANOVA by Sex and Observationin Writing at Suburban School (K-1)
Source
Between Subjects
Sex (A)
S/A
Within SubJects
Observation (B)
Sex X Observation (AB)
df. dfHIP dfERR P E.<
13
1 5.00 8.00 0.87 .54
12
28
2 10.00 40.00 2.95 .007
4.00 20.50 0.64 .64
2 10.00 40.00 0.96 .49
4.00 20.50 0.60 .67
SB /A 24
TOTAL 41
225
Table C.18
Standard Discriminant Function Coefficients and Univariate ANOVAson Use of Cohesion Categories in Writing for Observation at
Suburban School. (1C1)
Cohesion Category
DiscriminantFunction
CoefficientsUnivariate F Tests(2, 24)
Restricted Exophoric Reference -.584 .83 .45
Reference -.400 .16 .86
Ellipsis -.634 .89 .43
Conjunction -.560 4.03 .03
Lexical Cohesion -.985 10.74 .001
NM on is or or as or is um nit als um on se
Table C.19
Means and Standard Deviations of Cohesion Categoriesby Mode and Observation at the Suburban School (K-1)
Mode Observation R Exo Ref Reference Ellipsis Conjunction Lexical
Dictation
Writing
X SD X SD X SD X SD X SD
.01 .01 .42 .12 .01 .03 .21 .06 .37 .11
1 .01 .02 .42 .19 .01 .02 .21 .05 .38 .15
2 .01 .01 .38 .06 .01 .02 .22 .09 .40 .09
3 .00 .00 .47 .06 .02 .04 .20 .08 .34 .10
.03 .12 .48 .35 .05 .24 .10 .11 .29 .20
1 .03 .09 .52 .53 .01 .02 .04 .06 .13 .17
2 .06 .19 .44 .26 .12 V.42 .11 .13 .35 .20
3 .00 .00 .48 .19 .03 .07 .15 .12 .38 .15
ObservationOverall: 1 .02 .07 .47 .39 .01 .02 .12 .10 .26 .20
2 .03 .14 .41 .19 .07 .30 .17 .12 .37 .15
3 .00 .01 .47 .14 .02 .06 .17 .10 .36 .13
227
1
APPENDIX D
Story Structure MANOVAs in Retelling DictatiJn and Writing
at Urban (K-1) and Suburban Schools (K-1)
228
1-
APPENDIX D
Story Structure MANOVAs in Retelling Dictation and Writingat Urban (K-1) and Suburban Sch:.ols (X1)
Seven multivariate analyses of variance, followed up by discriminant
and univariate analyses, were employed to compare story structure data.
In the first of these MANOVAs, frequencies of function types, functions,
and moves for urban - school dictation and writing were organized into a
design where dialect and sex were between-subjects factors and where
mode and observation were within-subjects factors. Subsequent urban - school
MANOVAs analyzed story structure data for each mode separately using
essentially the same designs, with, of course, the mode factor eliminated as
a comparison. The suburban school MANOVAs for story structure data
followed the same pattern,without a dialect comparison. In the suburban-
school MANOVAs, sex was the between-subjects comparison, while mode and
observation served as within-subjects factors. In subsequent separate
comparisons for each mode, observation served as the 'ithin- subjects
factors.
The first MANOVA, comparing urban-school dictation and writing data,
resulted in a significant multivariate test statistic for the mode and
observation factors (See Table D.1 ). No significant first- and second -
order interactions were indicated by this analysis. Univariate follow-up
procedures resulted in significant test statistics for the mode factor on
function types, functions and moves (See Table D.2 ). Dictation means were
significantly higher than writing means (see Table D.4) for all three
dependent variables. Moves, made the strongest contribution w 'he
discriminant function followed by functions, then function types.
-191- 229
Follow-up procedures for the observation factor also indicated that
all three dimendent variables contributed about equally to the discriminant
function, and that all three increased significantly over observations.
Tukey ggaz h2g, comparisons of means (See Table D.4 ) for moves indicated
that number of moves increased significantly at observation two, but
declined slightly at observation three -- still, however, significantly
higher than observation one.
The story structure MANOVA for dictation at the urban schobl produced
a significant effect for the observation factor (See Table D.5 ). After
removal of the effects associated with the first root for the observation
factor, a significant discriminition remained for the second root. The
first root appeared to be composed mainly of variance associated with
function types while the second root appeared to be composed of variance
associated with functions and moves. All three story structure variables
achieved significant univariate test statistics fot the observation factor
(See Table D.6 ). Tukey per, laucomparisons for function types indicated
a significant increase in frequencies at observatir, two and a significant
decrease in frequencies at observation three. Function types followed
the same pattern. Moves increased significantly at observation two but
declined slightly at observation three but still significantly higher at
observation.three than at observation one. Means and standard deviations
are displayed in Table D.4.
No significant effects were indicated by the urban school MANOVA for
writing (See Table D.7 ).
The kindergarten MANOVA for story structure data in retelling produced
significant main effects for the observation fa tor, a significant second-
order dialect by sex by observation interaction, and a significant
23(J
-192-
1
1
1
1
first-order dialect by observation interaction. Follow-up discriminant
and univariate analyses of the dialect by sex by observation interaction
(See Table D.9 ) indicated no significant univariate effects for function
types, functions or moves. Therefore, the second order interaction was
probed with follow-up techniques (See Table D.10 ) for the dialect by
11
observation interaction indicating significant univariate test statistics
for function types and functions. (See Table D.10 ). Tukey max
comparisons for retelling means delonstrated that nonvernacular means
decreased sigoificantly at observation two and increased significantly
at observation three, while vernacular means for function types increased
significantly at observation two, then remained stable through first
grade. A similar interaction was indicated for functions (See Figure
D.1.). Means and standard deviations for dialect by observation are dis-
played in Table D.9. Tukey gm; h2g.comparisons of mean differences for
the functions interaction indicated that vernacular means increased
significantly at observation two, then declined with no significant
difference re-lining between observations one and three. Nonvernacular
11 means increased significantly at observation three producing differenceo
between observations one and three and two and three.
Observation factor follow-ups were not pursued given the significant
11
disordinal dialect by observation interaction.
The suburban- school story structure MANOVA for dictation and writing
11 produced no significant multivariate test statistics (See Tables D.11,
I/
D.12, and D.13 ). Only the story structure MANOVA for retelling produced
a significant multivariate effect--a significant test statistic for the
observation factor (See Table D.14 ). After effects associated with the
leading root for observation were removed, a significant discrimination
remained. The first root appeared to be composed primarily of variance
associated with moves, which declined significantly between observations
one and two and between observations one and three, and with function
types, which decreased significantly between observations one and two
but increased significantly between observations two and three. Function
types and functions contributed strongly to the second root --function
types and functions following very different patterns across observations
(See Table D.15). Means and standard deviations for suburban school (X-1)
story structure variables by observation are presented in Table D.16.
Table D.1
Story Structure MANOVA in Dictation and Writingby Dialect, Mode and Observation at Urban School (K-1)
Source df dfHYP dfERR F p. <
Between Subjects 9
Dialect (A) 1 3.00 6.00 .15 .92
S/A 8
Within Subjects 50
Mode (B) 1 3.00 6.00 20.00 .002
Dialect X Mode (AB) 1 3.00 6.00 1.22 .38
SB/A 8
Observation (C) 2 6.00 28.00 3.62 .009
2.00 14.50 2.35 .13
Dialect X Observation (AC) 2 6.00 28.00 .89 .51
2.00 14.50 .22 .80
SC/A 16
Mode X Observation (BC) 2 6.00 28.00 1.93 .11
2.00 14.50 1.66 .22
Dialect X Mode X 2 6.00 28.00 .38 .88
Observation (ABC) 2.00 14.50 .56 .58
SBC/A 16
TOTAL 59
233
Table D.2
Discriminant Function Coefficients and Univariate ANOVAs asFollow-up to Mode Effect for Urban School (K-1)
Criterion
Standard DiscriminantFunction Coefficients
Univariate F TestsF t.
Types 0.091 41.27 .001
Functions -0.359 37.13 .001
Moves 1.199 77.50 .001
Table D.3
Discriminant Function Coefficients and Univariate ANOVAs asFoliow-up to ObservationEffect for Urban School (-1)
CriterionStandard Discriminant/Unction Coefficients
Univariate F TestP p.4
Types 1.457 4.29 .03
Functions -1.616 4.80 .02
Moves 1.428 7.05 .006
234
Table D.4
Means and Standard Deviatiohs of Story Structure.Data for Urban School by Mode and Observation
Mode Observation Functions Types Moves
R SD X SD TE SD
Dictation 6.82 7.17 4.42 3.79 1.53 1.38
1 4.50 6.03 3.40 4.52 .95 1.052 10.25 8.11 6.35 3.27 1.95 1.283 5.70 6.17 3.50 2.76 1.70 1.63
Writing .77 1.91 .73 1.79 .25 .51
1 .20 .89 .20 .89 .05 .22
2 .85 2.01 .75 1.65 .30 .57
3 1.25 2.43 1.25 2.43 .40 .60
Observation
Means:1 2.35 3.43 1.80 2.78 .57 .98
2 5.55 6.10 3.55 1.89 1.12 1.093 3.97 4.24 2.37 2.51 1.05 1.03
Table D.5
Story Structure MANOVA in Dictation byDialect, Sex, and Observation -- for Urban School (X-1)
Source df df1170 dfERR F E.<
Between Subjects 19
Dialect (A) 1 .:.00 14.00 .72 .55
Sex (B) 1 3.00 14.00 1.64 .23
Dialect X Sex (AB) 1 3.00 14.00 1.38 .29
S/AB 1.6
Within Subjects 40
Observation (C) 2 6.00 60.00 3.05 .01
2.00 30.50 3.23 .05
Dialect X Observation (AC) 2 6.00 60.00 .85 .54
2.00 30.50 .49 .62
Sex X Observation (BC) 2 6.00 60.00 .94 .48
2.00 30.50 1.15 .33
4alect X Sex X 2 6.00 60.00 1.02 .42
Observation (ABC) 2.00 30.50 .80 .46
SC/AB 32
TOTAL 59
Table D.6
Discriminant Function Coefficients and Univariate ANOVAsas Follow-up to Observation Effect in Dictation for
Urban School (K-1) -- Story Structure
Criterion
Standard DiscriminantFunction Coefficients
Root 1 Root 2
Univariate F TestsF E. 4.
Tioes .917 .069 6.31 .005
FUnctions -.120 1.087 5.03 .01
Moves .315 -1.514 4.42 .02
I/
Table D.71
Story Structure MANOVA in Writing by Dialect,Sex, and Observation -- for Urban School (KA.)
Source df dfHYP dfERR 17.
Between Subjects 19
Dialect (A) 1 3.00 . 14.00 1.31 .31
Sex (B) 1 3.00 14.00 .53 .67
Dialect X Sex (AB) 1 3.00 14.00 .47 .71
SLAB 16
Within Subjects 40
Observation (C) 2 6.00 60.00 1.83 .11
2.00 30.50 1.37 .27
Dialect X Observation (AC) 2 6.00 60.00 1.24 .30
2.00 30.50 .70 .51
Sex X Observation (BC) 2 6.00 60.00 .45 .84
2.00 30.50 .34 .72
Dialect X Sex X 2 6.00 60.00 .47 .83
Observation (ABC) 2.00 30.50 .34 .71
SC/AB 32
TOTAL 59
-200-
1
II
11
1
II
11
1
238 1
11
Table D.8
Story Structure MANOVA in Retelling by Dialect,Sex, and Observation -- for Urban School (R-1)
Source df dfHYP dfERR 2..<
Between Subjects 19
Dialect (A) 1 3.00 14.00 1.52 .25
Sex (B) 1 3.00 14.00 .10 .96
Dialect X Sex (AB) 1 3.00 14.00 1.30 .31
sIAB 16
Within SubJects 40
Observation (C) 2 6.00 60.00 5.89 .001
2.00 30.50 .53 .59
Dialect X Observation (AC) 2 6.00 60.00 6.13 .001
2.00 30.50 1.74 .19.
Sex X Observation (BC) 2 6.00 60.00 2.05 .07
2.00 30.50 1.24 .30
Dialect X Sex X 2 6.00 60.00 2.35 .04
Observation (ABC) 2.00 30.50 .45 .64
SC/AB 32
TOTAL 59
Table D.9
Means and Standard Deviations for Function Types and Functionsat Urban School (K-1) by Dialect and Observation
Dialect Observation Function Types Functions
Vernacular
Nonvernacular
Overall
1
2
3
1
2
3
SD M SD
6.50 3.77 7.30 4.098.00 2.62 10.03 3.827.20 4.69 9.60 7.43
9.50 4.54 13.40 7.79
8.10 2.71 10.90 5.2113.30 4.79 19.40 8.97
8.77 4.53 11.82 7.47
1 8.00 4.26 10.35 6.682 8.05 3.25 10.60 5.103 10.25 5.63 14.50 9.55
240
Table D.10
Discriminant Function Coefficients and Univariate ANOVAsas Follow-up to Dialect by Observation Interaction inRetelling for Urban School(K-1)-- Story Structure
CriterionStandard DiscriminantFunction Coefficients
Univariate F TestsP 2. 4
Types 0.704 9,48 .001
Functions 1.026 5,28 .01
Moves -1.392 1.81 .18
2
18
16
14
12
1
08
0
02
0301
FUNCTION TYPES
03
0102
2'
18
16
14
12
1
08
06
04 04
02 02
0 0
02
03
01
FUNCTIONS
03
01
02
Vernacular Nonvernacular Vernacular Nonvernacular
Figure D.I. Observation as a Functionof Dialect at Urban School (K-1)for Function Types an' Functions
Table D.11
Story Structure MANOVA by Sex, Mode, andObservation -- for suburban School (K-1) -- Dictation and Writing
Source df dfHYP dfERR F p,. <
Between Subjects 9
Sex (A) 1 3.00 6.00 1.32 .35
S/A 8
Within Subjects 50
Made (B) 1 3.00 6.00 3.46 .09
Sex X Mode (AB) 1 3.00 6.00 3.75 .08
SB/A 8
Observation (C) 2 6.00 28.00 .91 .50
2.00 14.50 .12 .89
Sex X Observation (AC) 2 6.00 28.00 .49 .81
2.00 14.50 .06 .94
SC/A 16
Mode X Observation (BC) 2 6.00 28.00 1.37 .26
2.00 14.50 .82 .46
Sex X Mode X 2 6.00 28.00 .52 .79
Observation (ABC) 2.00 14.50 .35 .71
SBC/A
TOTAL
Table D.12
Story Structure MANOVA in Dictation bySex and Observation -- for Suburban School (K-1)
Source df dfHYP dfERR 7 E. 4
Between Subjects 9
Sex (A) 1 3.00 6.00 3.47 .09
S/A 8
Within Subjects 20
Observation (B) 2 6.00 28.00 .42 .86
2.00 14.50 .21 .81
Sex b Observation (AB) 2 6.00 28.00 .52 .79
2.00 14.50 .18 .84
SB/A. 16
TOTAL 29
Table D.13
Story Structure MANOVA in Writing bySex and Observation -- Suburban School (K-1)
Source df dfHYP dfEB11 F g. <
.111111,!,...
Between SOlects, 9
Sex (A) 1 3.00 6.00 1.17 .40
S/A 8
Within Subjects 20
Observation (B) 2 6.00 28.00 1.92 .34
2.00 14,50 .31 .74
Sex X Observation (AB) 2 6.00 28.00 .79 .58
2.00 14.50 .15 .86
SB/A 16
TOTAL 29
245
-207-
Story StructurA MANOVA in Retelling by Sexand Observation -- for Suburban School (K-l)
Source df dfHYP dfEart F 2..4=1,
Be..11Subects 9
Sex (A) 1 3.00 6.00 1.80 .25
S/A 8
WithinAgkitla 20
Observation (B) 2 6.00 28.00 8.49 .001
2.00 14.50 3.63 .05
Sex X Observation (AB) 2 6.00 28.00 2.36 .06
2.00 14.50 1.53 .25
SIVA 16
TOTAL 29
246-208-
Table D.15
Discriminant Function Coefficients and Univariate ANOVAs arFollow-up to Observation Effect in Retelling for Suburban
School (K-1) -- Story Structure
CriterionStandard DiscriminantAInction CoefficientsRoot i Root 2
Univariate F TestsF 2., 4
Types 1.091 -2.044 4.13 .04
Functions 0.921 2.787 2.06 .16
)hues -2.066 -0.552 5.70 .01
Table D.16
Means and Standard Deviations of StoryStructure Data in Retelling for Suburban
School by Observation (K,..1)
Observation Functions Types Moves
X SD 3E SD 3E SD
1 16.20 5.87 12.00 4.03 2.40 .70
2 12.70 5.89 8.80 3.19 1.70 .48
3 17.10 6.59 13.20 5.05 1.50 .53
APPENDIX E
Text Length and Syntactic Complexity
Text length and syntactic complexity have been employed in a
variety of language acquisition studies as convenient measures to
equate language samples being studied. Both of these measures have
been the subjects of considerable discussion regarding their meaning,
utility, and reliability. But, because utterance length and syntactic
complexity have been so widely utilized, they have become benchmarks
to the field, and, accordingly, length of dictated and written texts,
measured by number of T-Units, and syntactic complexity of these texts,
measured by mean T-Unit length, have been analyzed to provide a basis
for comparison with other language development research and earlier
writing research. While we question whether text length and mean
utterance length shed much light on writing development, these measures
and comparisons may be of interest to other researchers in the field.
Text Length: Dictation and writing
Three separate ANOVAs were employed to compare length of dictated
and written texts produced by: (I) kindergarten children in both
urban and suburban schools, (2) kindergarten children in the urban
school, and (3) kindergarten children in the suburban school. Three
additional ANOVAs were employed to compare length of dictated texts
produced by: (1) kindergarten children in both schools, (2) kinder-
garten children in the urban school, and (3) kindergarten children in
the suburban school. Conservative F tests were employed for all
significant effects involving within-subjects factors, and for all
Tukey follow-up comparisons.
School comparisons: School and sex served as between-subjects
factors while mode and observations served as within-subjects factors
-212- 25u
for this comparison. Table E.1 presents means and standard deviations
of text length for dictation and writing produced by kindergarten
children at both schools. Significant test statistics were obtained
Table E,1
Means and Standard Deviations of Text. Length in Dictationand Writing for both Schools (1C.1) -- by Mode and Observation
Mode Observation Mean SD
Dictation 22.22 26.40
1 12.00 10.762 25.60 32.113 29.05 29.10
Writing 6.32 6.15
1 2.45 2.09
2 6.05 5.203 10.45 7.24
ObservationOverall; 1 7.23 9.05
2 13.83 24.77
3 19.75 22.95
for the observation and mode factors (See Table E.2). As might be
expected, dictated texts were significantly longer than written texts.
Tukey post-hoc tests indicated that texts were significantly longer
at each observation.
-213-
Table E.2
ANOVA of Text Length in Dictation and Writingby School, Sex, Mode, and Observation (K-1)
Source df MS E. <
Between Subjects 19
School (A) 1 952.05 3.00 .10
Sex (B) 1 172.78 .54 .47
School X Sex (AB) 1 17.64 .06 .82
S/AB 16 317.75
Within Subjects 100
Node (C) 1 7584.09 22.58* .001
School X Node (AC) 1 770.10 2.29 .15
Sex X Mode (BC) 1 20.83 .06 .81
School X Sex X Mode (ABC) 1 258.12 .77 .39
SC/AB 16 335.89
Observation (D) 2 1641.55 5.62* .008School X Observation (AD) 2 594.61 2.04 .15
Sex X Observation (BD) 2 617.54 2.12 .14
School X Sex XObservation (ABL) 2 1038.78 3.56** .04
SD/AB 32 291.99
Node X Observation (CD) 2 304.65 1.05 .36
School % Mode XObservation (ACD) 2 578.58 1.99 .15
Sex X MOde XObservation (BCD) 2 341.08 1.18 .32
School X Sex X Mode XObservation (ABCD) 2 921.77 3.18 .06
SCD/AB 32 290.11
TOTAL 119
* Geisser -Greenhouse conservative F test using reduced degrees of freedom(1, 16) was significant at 2. <.05.
** Geisser -Greenhouse conservative F test using reduced degrees of freedom(1, 32) was not significant at 2. <.05.
-214-252
Urban school: In the urban school ANOVA for text length, dialect
. and sex were between subjects factors and mode and observation were
within-subjects factors. Table E.3 presents means and standard
Table E.3
Means and Standard Deviations of Text Length in Dictation andWriting at Ur ban School (K-1)--by Sex, Mode and Observation
Mode Sex Observation 3E S.D.
Dictation 27.93 31.13
Male 27.53 32.871 17.80 19.712 16.50 12.49
3 48.30 47.08
Female 28.33 29.851 12.70 13.912 48.10 42.213 24.20 13.07
Writing 4:57 6.59
Male 4.13 6.261 0.60 1.082 3.60 6.423 8.20 7.18
Female 5.00 6.981 2.20 2.532 4.50 7.093 8.30 8.91
SexOverall:
Males1 9.20 10.402 10.05 9.463 28.25 24.08
Females1 7.45 8.222 26.30 24.653 16.25 10.99
ObservationOverall: 1 8.33 9.44
2 18.18 20.48
3 22.25 21.85
253-215-
1
1
1
1
1
1
1
deviations for text length by mode, sex, and observation. Significant
test statistics were obtained for a second-order, sex by mode by
observation interaction, for a first-order, sex by observation inter-
action, and for mode and observation main effects (See Table E.4).
Table E.4
ANOVA of Text Length in Dictation and Writing atthe Urban School -- by Dialect, Sex, Mode, and Observation
Source df MS F k.<
Between Subjects 19
Dialect (A) 1 83.32 .15 .71
Sex (B) 1 20.83 .04 .85
Dialect X Sex (AB) 1 19.20 .03 .86
S/AB 16 561.67
Within Subjects 100
Mode (C) 1 16379.65 27.58* .001
Dialect X Mode (AC) 1 172.79 .29 .60
Sex X Mode (BC) 1 .03 .00 .99
Dialect X Sex X Mode (ABC) 1 128.12 .22 .65
SC/AB 16 593.87
Observation (D) 2 2050.17 5.06* .01
Dialect X Observation (AD) 2 381.25 .94 .40Sex X Observation (BD) 2 2045.11 5.05* .01
Dialect X Sex XObservation (ABD) 2 129.17 .32 .73
SD/AB 32 405.14
Mode X Observation (CD) 2 679.39 1.70 .20
Dialect X Mode XObservation (ACD) 2 270.46 .68 .52
Sex X Mode XObe'rvation (BCD) 2 1966.16 4.92* .01
Dialect X Sex X Mode XObservation (ABCD) 2 18.26 .05 .96
SCD/AB 32 399.37
TOTAL 119
* Geisser-Greenhouse conservative I-test using reduced degrees of freedom(1, 16) was significant at 2. 4.05.
-216- 254
The significant second-order, sex by mode by observation interaction
is graphed in Figure E.1. Females dictated significantly longer texts
at observation 2 than at observation one or three while males dictated
5048464442403836343230282624222018161412100806040200
01
02
2019181716
1514131211
03 100908 03 03
070605 0
204 020302 01
01
0001
MALE FEMALE
DICTATION3028 0326242220181614
1210
0806040200
02
MALE FEMALE
WRITING
, 01
MALE FEMALE
Figure E.1 Observation as a Function of Sex and Mode on Text Length for Urban
'School (K-1)
-217-255
significantly longer texts at observation 3 than at observation one
or two. Tukey post hoc tests indicated that girls dictated signifi-
cantly longer texts than boys at observation 2 while boys dictated
significantly longer texts at observation three than did girls. As
shown in Figure E.1, overall girls produced significantly longer
texts at observation 2 than they did at observation 1. Girls' texts
at observation 2 were significantly longer than boys' texts at either
observation 1 or observation 2. Boys produced significantly longer
texts at observation 3 than at observation 2, and significantly longer
texts at observation 2 than they did at observation 1. Tukey follow-up
procedures for the mode effect indicated that dictated texts (11".27.93)
were significantly longer than written texts (1+ 4.57). Because the
pattern for text length across the three observations was not the same
for boys and girls, no attempt was made to interpret the main effect
for observation.
Suburban school: Design arrangements for the suburban ANOVA on
dictation and writing were similar to school and urban comparisons
noted above: sex served as a between-subjects factor while mode and
observation served as within-subjects factors. Table E.5 presents
means and standard deviations for text length in the suburban kinder-
garten texts. Table E.6 summarizes the ANOVA results-main effects for
sex and mode, and a significant first-order, sex X mode interaction.
The sex X mode interaction (See Figure E.2) indicated that girls dictated
significantly longer texts than boys but wrote texts of roughly the same
length that boys wrote. Because the patterns of text length differed
for boys and girls in dictation, no attempt will be made to interpret
the main effect for sex, but Tukey post hoc tests indicated that dictated
texts were significantly longer than written texts for both sexes, thus
-218- 256
accounting for the significant code factor effect.
Table E.5
Mears and Standard Deviations of Text Length inDictation and Writing by Sex, Mode, and Observation
at Suburban School (K-1)
Sex Mode Observation R SD
Males 9.82 6.76
Dictation 13.40 6.91
1 15.00 10.842 14.00 5.703 11.20 2.68
Writing 6.33 4.50
1 2.80 1.302 6.40 2.973 9.80 5.45
Females 13.03 12.28
Dictation 20.33 13.48
1 15.60 9.24
2 16.20 11.843 29.20 16.21
Writing 5.73 4.06
1 2.00 2.352 7.60 2.88
3 7.60 4.22
''.de
,1: Dictation: 16.87 11.10Writing: 6.03 4.22
(..,servation
Overall: 1 8.85 9.392 /1.05 7.623 14.45 12.04
1
1
1
1
Table E.6
ANOVA of Text Length in Dictation and Writing at theSuburban School by Sex, Mode, and Observation (K-1)
Source df MS F 2.4
Between Subjects 13
Sex (A) 1 192.01 4.77 .05
S/A 12 40.26
Within Subjects 70
Mode (B) 1 3132.92 71.63* .001
Sex Mode (AB) 1 398.67 9.12 .01
SB/A 12 43.74
Observation (C) 2 297.74 '4.41** .02
Sex X Observation (AC) 2 57.44 .85 .44
SC/A 24 67.51
Mode X Observation (BC) 2 72.75 .93 .41
Sex X Mode XObservation (ABC) 2 122.24 1.55 .23
SBC/A 24 78.66
TOTAL 83
IMIM
* Geisser-Greenhouse conservative F test using reduced degrees offreedom (1,12) was significant at z. < .05.
** Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 12) was not significant at 2.4c.05.
258-220-
2422
2018
16
14
12
10
0806
04
02
00
DICTATION
WRITINGVAirt *la = Immo 11.10
MALE FEMALE
Figure E.2 Mode as a Function of Sex on Text Length for Suburban School(Kr.1)
Text Length in Dictation Onl,,
Three additional ANOVAs were employed to analyze the text length
variable ill dictation only, again comparing schools, sex and dialect as
between- subjects factors and observation'as a within - subjects factor in
the three respective designs. No significant main or interaction effects
were obtained on the school ANOVA (See Table E.7), but for the urban
school ANOVA, a significant first-order, sex X observation effect was
indicated (See Table E.8), an interaction graphed earlier in Figure E.1,
Means and standard deviations were presented in Table E.3. Tukey post-
hoc comparisons indicated that girls dictated longer texts at observation
two than at one or three while males dictated significantly longer
texts at observation three than at one or two. girls dictated signifi-
cantly longer texts than boys at observation two while boys dictated
significantly longer ,exts at observation three than did girls. For the
suburban school ANOVA, a main effect for sex was indicated (Sec Table E.9).
253
Table E.7
ANOVA of Text Length in Dictation by School,
Sex, and Observation (X-1)
Source df MS F P. <
Between Subjects 19
School (A) I, 1717.36 2.78 .12
Sex (B) I, 156.82 .25 .62
School X Sex (A3) I, 205.35 .33 .57
S/AB 16 617.50
Within Subjects
Observation (C) 2 1625.18 2.91 .07
School X Observation (AC) 2 1135.93 2.04 .15
Sex X Observation (BC) 2 932.18 1.67 .20
School, X Sex X
Observation (ABC) 2 1958.72 3.51* .04
SC/AB 32 558.07
TOTAL
* Geisser -Greenhouse conservative F test using reduced degrees offreedom (1, 32) was not significant at 1.4 .05.
I
Table E.8
ANOVA of Text Length in Dictation by Dialect,Sex, and Observation at the Urban School (K-1)
Source df MS
Between Subjects 19
Dialect (A) 1 8.06 .007 .93
Sex (B) 1 9.60 .01 .93
Dialect X Sex (AB) 1 24.07 .02 .89
S /AB 16 1117.41
Within Subjects 40
Observation (C) 2 2490.99 3.21* .05
Dialect X Observation (AC) 2 596.60 .77 .47
Sex X Observation (BC)bialect X Sex X
2 4008.56 5.17** .01
Observation (ABC) 2 114.52 .15 .86
SC/AB 32 776.05
TOTAL 59
* Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 16) was not si3nificant at z 4.50.
**Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 16) was significant at 2.. 4.05.
-223-
26i
1
II
II
I
11
Table E.9
ANOiA of Text Length in Dictation at SuburbanSchool (K-1) by Sex and Observation
Source df MS
Between Subjects 13
Sex (A) 1 572.02 8.83 .01
S/A 12 64.79
Within Sub ects
Observation (B) 2 220.50 1.56 .23
Sex X Observation (AB) 2 169.30 1.20 .32
SB/A 24 141.15
TOTAL
Means and szandard deviations for text length in dictation were pre-
sented in Table E.5. Tukey post hoc comparisons indicated that girls
dictated longer texts than boys at the suburban school.
Syntactic Complexity: Dictation and Writing
Three separate ANOVAs were employed to compare mean T-unit length
for dictated and written texts produced by kindergarten children as
above at both the urban and suburban schools and within each school.
School compari7on: School and sex served as between-subjects
factors; mode and observation served as within-subjects factors In
this comparison. This analysis produced five significant effects
(See Table E.10): school, mode and observation factors plus a second-
order interaction for school by sex by observation and a first-order
-224- 262...11
Table 8.10
ANOVA of Syntactic Complexity in Dictation andLing by School, Sex, Mode, and Observation (K-1 Population)
Source
Between Subjects 19
School (A) 1 50.84 6.00 .03
Sex (B) 1 4.05 .48 .50
School X Sex (AB) 1 22.01 2.60 .13
SIAB 16 8.48
Within Sublects 100
Made (C) 1 278.92 43.34* .001
School X Mode (AC) 1 5.71 .89 .36
Sex X Made (BC) 1 .01 .00 .97
School X Sex X Mode (LBC) 1 .11 .02 .90
SC/AB 16 6.44
Observation (D) 2 21.34 6.02 .006
School X Observation (AD) 2 .98 .28 .76
Sex X Observation 2 2.83 .80 .46
School X Sex XObservation (ABD) 2 16.01 4.51** .02
SD/AB 32 3.55
Made X Observation (CD) 2 34.62 10.15** .001
School X Mode XObservation (ACD) 2 11.09 3.25*** .05
Sex X Mode XObservation (BCD) 2 2.60 .76 .48
School X Sex X Mode XObservation (ABCD) 2 1.37 .40 .67
*CD/AB 32 3.41
TOTAL 119
* Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 16) was significant at 2.. < .05.
** Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 16) was not significant atil.< .05.
***Geisser-Greenhouse conservative F test using reduced degrees offreedom (1, 16) was not significant at E. 4.05.
263-225-
interaction for mode by observation. Note that the Geisser-Greenhouse
conservative F test for the school by mode by observation interaction
failed to reach significance.
Table E.11
Means and Standard Deviations of Syntactic Complexity inDictation and Writing by School, Sex and Observation (IC-1)
School Sex Observation Mean S.D.
Urban 5.68 3.29Males 5.07 3.22
1 4.03 1.49
2 4.91 3.433 6.27 1.61
Females 6.29 3.29
1 6.91 2.402 5.07 2.633 6.90 2.57
Suburban 6.98 2.32Males 7.23 1.80
1 7.39 1.812 6.97 0.853 7.32 0.94
Females 6.74 2.75
1 5.70 2.632 6.33 0.943 8.19 2.47
SexOverall
Hales1 5.71 1.652 5.94 2.193 6.80 1.28
Females1 6.31 2.522 5.70 1.79
3 7.55 2.52
Tukey post-hoc tests, comparing the means (See Table E.11) of
the school by sex by observation interaction indicated that urban-
school girls produced texts of greater syntactic complexity than urban
8.5'
8.00
7.507.006.506.005.50
03 --0-'"....uniug-
A/,01
440 03
8.508.007,507,006.506.005.50
5.00 02 02 5.00
4.50 4.50
4.00 01 4.00
3.50 3.50
3.00 3.002.50 2.502.00 2.00
1.50 1.50
1.00 1.000.50 0.500.00 0.00
03
02
01
MALE FEMALE MALE FEMALE
URBAN SUBURBAN
8.508.007.507.006.506.005.505.004.504.00
3.50
3.00
LAO2 00
1.501.00
0.500.00
033
02 01
01 02
MALE FEMALE
SCHOOL
Figure E.3. Observation as a Function of Sex and School in Dictationand Writing (K-1)
-227-
265
boys at observation one while the converse held at the suburban
school: boys produced texts of greater syntactic complexity than
girls. But, at observation three, suburban girls produced texts of
greater syntactic complexity than suburban boys (See Figure E.3).
The disordinal character of this interaction is further indicated by
the significant decrease in syntactic complexity for urban girls
between observations one and two followed by a significant increase
between observations two and three with no significant net gain in
syntactic complexity over observations. On the other hand, syntactic
complexity means for urban boys increased significantly between
observations one and three. Conversely, at the suburban school,
syntactic complexity means for boys did not vary significantly from
observation to observation while for suburban girls, syntactic complex
ity means increased significantly between observations one and three
and two and three. The patterns indicated by this interaction are so
mixed that interpretations for both observation school factors are
too risky to attempt.
For the mode t; observation interaction, Tukey post hoc comparisons
of means (See Table E.12) indicated that while writing means for syntactic
complexity increased significantly at each observation, dictation means
declined significantly between observations one and two, only to
increase significantly between observations two and three to a level
commensurate with observation one (See Figure E.4). Syntactic complexity
means for dictation were significantly higher than for writing at every
observation. The significant main effect for mode, therefore, indicated
that dictated texts were significantly more complex syntactically than
were written texts (See Table E.12).
Table E.12
Means and Standard Deviations of Syntactic Complexity (K- 4) - -by
Mode and Observation
Made Observation 1L S.D.
Dictation1 8.61 2.302 6.80 1.813 8.17 1.77
Writing1 3.41 2.492 4.84 2.093 6.18 2.31
Made Dictation 7.86 2.22Overall Writing 4.81 2.72
10
09DICTATION
080706OS WRITING04
dm+40+
0302
01
00
01 02 03
Figure E.4. Mode as a Function of Observation in Dictation and Writing(K-1)
Urban school comparison: Dialect and sex served as between -
subjects factors while mode and observation served as within-subjects
factors in this analysis of syntactic complexity in children's dictated
and written texts. This analysis produced two second-order interactions,
dialect by sex and mode by observation, and significant effects for
the mode and observation factors (See Table E.13).
-229-267
Sable E.13
ANOVA of Syntactic Complexity in Dictation and Writing
at the Urban School (K-1) by Dialect, Sex, Mode, and Observation
Source df MS P R.<
Between Subjects 19
Dialect (A) 1 3.22 .38 .54
Sex (B) 1 .00 .00 .99
Dialect X Sex (AB) 1 44.66 5.33 .04
SlAB 16 8.39
Within Subjects 100
Mode (C) 1 556.71 61.10* .001
Dialect X Mode (AC) 1 20.29 2.23 .16
Sex X Mode (BC) 1 .56 .06 .81
Dialect X Sex X Mode (ABC) 1 .26 .03 .87
SC/AB 16 9.11
Observation (D) 2 28.05 4.90* .01
Dialect X Observation (AD) 2 5.50 .96 .39
Sex X Observation (BD) 2 3.93 .69 .51
Dialect X Sex XObservation (AID) 2 9.09 1.59 .22
SD/AB 32 5.73
Mode X Observation (CD) 2 56.04 12.37 . .001
Dialect X Mode XObservation (ACD) 2 14.63 3.23** .05
Sex X Mode XObservation (BCD) 2 .19 .04 .96
Dialect X Sex X Mode XObservation (ABCD) 2 2.87 .63 .54
SCD/AB 32 4.53
Wen 119
* Ceisser-Greenhouse conservative F test using reduced degrees of freedom
(1, 16) was significant at p,. 6.05.** Geisser-Greenhouse conservative F test using reduced degrees of freedom
(1, 16) was not significant at E. 4.05.
-230- 268
For the dialect by sex interaction, Tukey post hoc comparisons
(See Table E.14) indicated that syntactic complexity means were sig
nificantly higher for nonveracular girls than for vernacular girls,
which, indeed, was the only significant difference produced in this
set of post hoc comparisons (See Figure E.5).
Table E.14
Means and Standard Deviations of Syntactic Complexity inDictation and Writing at Urban School (K.1) by Dialect and Sex
Dialect Sex Mean S.D.
Vernacular
Nonvernacular
MaleFemale
MaleFemale
5.96 2.704.75 2.49
5.07 2.696.29 2.83
080706
0504.03
02
01
00
NONVERNACULAR
" ...""
VERNACULAR
MALE FEMALE
Figure 145. Dialect as a Function of Sex in Dictation and Writingat Urban School (K-1)
Similar post hoc comparisons of syntactic complexity means (See
Table E.15) for the significant mode by observation interaction showed
that syntactic complexity of dictated texts.decreased significantly
1
between observation one and observation two, but did not differ signifi-
cantly between observations two and three or between observations one and
three. Syntactic complexity in written testsm however, increased signifi-
cantly between each observation and over all observations (See Figure E.6),
but despite these increases, wr..ting means for syntactic complexity
remained significantly lower than dictation means (See Table E.15).
Table E.15
Means and Standard Deviations of Syntactic Complexity inDictation and Writing at Urban School (K-1) by Observation
Mode Observation Mean S.D.
Dictation1 8.33 2.51
2 7.15 2.523 7.54 2.14
Writing1 1,50 2.162 3.19 2.953 5.41 2.99
ObservationOverall
1 4.92 2.34
2 5.17 2.743 6.48 2.57
Mode Dictation 7.67 2.39Overall; Writing 3.37 2.70
10
090807060504030201
00
DICTATION
4,e
WRITING
01 02 03
Figure E.6. Mode as aFunction of Observation:at Urban School (K-1)
-232-27n
Because syntactic complexi.y patterns for dialect, sex, and mode, inter-
acted disordinally, no attempt will be made to interpret significant
effects for the mode and observation factors.
Suburban school comparison; Design arrangements for thic. comparison
were similar to those given above, the difference being the absense of a
Table E.16
ANOVA of Syntactic Complexity in Dictation andWriting at Suburban School by Sex, Mode, and Observation.
(K -l}
Source df ms F.2. 4
Betwe n Subjects 13
Sex (A) 1 5.46 .68 .42
S/A 12 7.99
Within Subjects 70
Mode (B) 1 .14.99 43.87* .001
Sex X Mode (AB) 1 1.99 .76 .40
SBA 12 2.62
Observation (C) 2 10.34 4.51** .02
Sex X Observation (AC) 2 7.69 3.36** .05
SC /A 24 2.29
Mode X Observation (BC) 2 3.68 2.00 diSex X Mode X
Observation (ABC) 2 6.31 3.42** .05
SBC /A 24 1.85
TOTAL 83
* Geisset-Greenhouse conservative F test using reduceraiiFiiia------freedom (1, 12) was significant at E. <.05.
** leisser-Greenhouse conservative F test using reduced degrees offreedom (1, 12) was not significant at jz.4 .05.
2 71
vernacular population at the suburban school, leaving just sex as a
between-subjects factor with mode and observation remaining as within-
subjects factors. Only a significant test statistic for mode was
obtained (See Table E.16). Note that the conservative F tests for the
Table E.11
'leans and Standard Deviations of SyntacticComplexity in Dictation and Writing by Sex, Mode, and
Observation at Suburban School (1(i)
Sex Mode Observation SD
Males Dictation 8.50 1.43
1 8.77 1.922 8.36 .75
3 8.36 1.66
Writing 5.95 1.09
1 6.02 1.712 5.59 .94
3 6.27 .22
Females Dictation 8.08 1.70
1 7.87 2.012 7.0 .59
3 9.28 1.63
Writing 5.39 2.99
1 3.53 3.252 5.55 1.293 7.10 3.31
Mode Overall:
Dictation: 8.29 1.56Writing: 5.68 2.23
Observation Overall: 1: 6.55 2.95
2: 6.65 1.473: 7.75 2.20
observation factor, the first-order sex by observation interaction, and
the second-c-der sex by mode X observation interaction failed to reach
significance. Tukey post hoc comparison of means for syntactic com-
plexity were significantly higher for dictation than they were for
writing (See Table E.17).
Summary: Text Length and Syntactic Complexity
Children wrote longer texts at each observation regardless of
school, sex, or dialect. The picture is not as clear for dictation.
In dictation no stable pattern for text length was discernable, owing
largely to opposing fluctuations in mean for boys and girls over
observations both between and within schools. Typically, dictated texts
decreased in length at the outset of second grade for boys but increased
for girls. Yet, by the end of second grade at the urban school, boys
and girls did not differ in the length of texts they dictated; at the
suburban school they did with girls dictating lon8er texts than boys.
Since the F values for interactions involving sex were usually marginal,
interpretation of this urban-suburban reversal between boys and girls
was too risky to attempt. Then too, variability in all calls w; ex-
ceptionally high which further underscored the futility of trying to
interpret dictation differences for tem:: length,
Syntactic complexity was defined as mean number of words per T- unit.
This measure as included as a dependent variable to provide other
researchers a well-known basis for characterizing language development
and for comparing these texts with those obtained In other studies
of language development.
In general, syntactic complexity means were higher for dictation
than for writing. Across observations, mean T-unit length in dictation
did not increase significantly; however, in writing mean T unit
length did increase significantly across observations. Again, a
complex pattern of interactio=s characterized the differences between
boys and girls for syntactic complexity. There was simply no basis
for a reasonable interpretation of these patterns.
274
236